+ All Categories
Transcript
Page 1: Rate-Distortion and Rate-Energy-Distortion Evaluations of ...

Research ArticleRate-Distortion and Rate-Energy-Distortion Evaluations ofCompressive-Sensing Video Coding

Bingyu Ji Ran Li and ChanganWu

School of Computer and Information Technology Xinyang Normal University Xinyang 464000 China

Correspondence should be addressed to Ran Li liran358163com

Received 21 December 2016 Accepted 7 March 2017 Published 16 March 2017

Academic Editor Jintao Wang

Copyright copy 2017 Bingyu Ji et alThis is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Compressive-Sensing Video Coding (CSVC) is a new video coding framework based on compressive-sensing (CS) theory Thispaper presents the evaluations on rate-distortion performance and rate-energy-distortion performance of CSVC by comparingit with the popular hybrid video coding standard H264 and distributed video coding (DVC) system DISCOVER Experimentalresults show that CSVC achieves a poor rate-distortion performance when compared with H264 and DISCOVER but its rate-energy-distortion performance has a distinct advantage moreover its energy consumption of coding is approximately invariantregardless of reconstruction quality It can be concluded that with a limited energy budget CSVC outperforms H264 andDISCOVER but its rate-distortion performance still needs improvement

1 Introduction

Video communication is an important type of data commu-nication Compression coding must be done before high-dimensional video signal is transmitted in channels withlimited bandwidth Therefore video compression coding hasbecome a hot research topic in digital video communicationThe international video coding standard H264 [1] jointlydeveloped by ISOIEC and the ITU-T has been widely usedin various video technologies and H264 has achieved greatcommercial success H264 standard uses motion estima-tion and discrete cosine transform to eliminate temporaland spatial redundancy of video sequences and its codingcomplexity is much greater than decoding complexity Forinstance when the test sequence Foreman with CIF formatis processed by H264 codec the encoding time is about 50 to90 times as long as the decoding time in different quantizationsteps whichmeans that H264 has strong applicability for thesituation of one coding and multiple decoding such as videobroadcasting and video on demand For the wireless com-munication equipment long-time encoding means reducedeconomics and practicality therefore video coding methodwith low coding complexity is needed as an alternative Inthis case DVC [2] which was first proposed by Wyner

and Ziv in information coding theory [3] has receivedwidespread attention In the initial stage of DVC researchthe main codec algorithms include Wyner-Ziv video coding[4] PRISM video coding [5] hierarchical Wyner-Ziv videocoding [6] and DVC scheme based on wavelet coding [7]With an aim of improving coding performance EuropeanUnion scientific research institutions put forward specialresearch plan and based on the existing research developa DVC standard program called DISCOVER (DIStributedCOding for Video sERvices) [8] DISCOVERmakes the low-complexity video coding performance further enhanced Butthe feedback channel [9] and the virtual channel [10] inDISCOVER scheme are highly controversial which is animportant engineering problem hindering its popularizationand application CS theory [11ndash13] combined with videocoding has led to the emergence of a new low-complexityvideo coding scheme called CSVC [14] The scheme stillretains the distributed characteristics and does not depend onfeedback channel or virtual channel and has great engineer-ing application potential which has attracted many scholarsrsquoattention [15ndash17]

At present there is still a lack of discussion on thecomparison of rate-distortion performance between CSVC

HindawiInternational Journal of Digital Multimedia BroadcastingVolume 2017 Article ID 4589124 8 pageshttpsdoiorg10115520174589124

2 International Journal of Digital Multimedia Broadcasting

MC

Intrapredictionselection

Intraprediction

ME

Inter

Intra NAL

Entropy coding

Reordering

Filtering

Current block Fn

Reference block F㰀

nminus1

Reconstructed block F㰀

n

Dn

T Q

X

P

D㰀n

Tminus1 Q

minus1

minus

+

uF㰀n

Figure 1 H264 codec architecture

H264 and DISCOVER It will help to clarify the upperlimit of performance improvement of CSVC by obtaining theresult of the performance difference between the three videocoding schemes The rate-distortion performance cannotshow the relationship between coding energy consumptionand video reconstruction quality Therefore rate-energy-distortion performance [18] also needs an objective eval-uation of CSVC H264 and DISCOVER This paper firstsummarizes the basic framework and technical details ofH264 and DISCOVER Next a typical algorithm of CSVCwill be described in detail Finally on the basis of the theoriesexposition rate-distortion performance and rate-energy-distortion performance of the three video coding schemesare evaluated and the performance difference between thethree video coding schemes is then fully discussed Theexperimental results show that under the test of CIF videosrespectively named Bus Football Foreman and Mobile interms of the rate-distortion performance H264 is optimaland DISCOVER follows while CSVC is the worst and hasa large performance difference from the other two but interms of rate-energy-distortion performance CSVC is opti-mal DISCOVER follows and H264 is the worst The resultsalso reveal a fact that the energy consumption of CSVC isapproximately the same regardless of reconstruction qualitywhile for H264 and DISCOVER there is a close correlationbetween recovery quality and the energy consumption ofcoding As a result it can be concluded that when low energyconsumption is demanded CSVC program can give full playto its advantages but its rate-distortion performance stillneeds improvement

2 Typical Video Coding Schemes

21 H264 H264 system is divided into two levels in func-tion Video Coding Layer (VCL) and Network AbstractionLayer (NAL) Its codec framework is shown in Figure 1 In thecoding process there are two options to predict the currentimage block F119899 interprediction and intraframe predictionWhen interprediction is adopted the motion vector of cur-rent block F119899 is obtained by the motion estimation accordingto the reference block and then the predicted frame Pcould be obtained by the motion compensation method

when intraprediction is used the predicted block in thecurrent frame is the weighted average of the selected adjacentdecoded blocks of the current block After the predictiveframe is determined the main steps of codec process are asfollows

Step 1 Calculate the residual D119899 between the current blockF119899 and the predicted value P

Step 2 Obtain the quantized coefficient 119883 by transformingand quantizingD119899

Step 3 Form the bit stream by reordering and entropy codingof quantized coefficient119883Step 4 Transmit the bit stream to the decoder side throughNAL at the same time use part of the bit stream which couldbe decoded on the encoder side as reference frame

The core technology of H264 is mainly reflected on itsimprovement in the interframe and intraframe predictivecoding for example using 4 times 4 integer discrete cosinetransform technology instead of the former 8 times 8 discretecosine transform technology to avoid mismatches in inversetransformation

22 DISCOVER The codec framework of DISCOVER sys-tem is shown in Figure 2 The encoder side carries on videoprocessing in Group of Pictures (GOP) whose length isdetermined by the specific situation The length of GOP willbe increased to reduce time redundancy when the imagecontains a small amount of motion on the other hand thelength can be shortened accordingly if there is a large amountof motion For each group video frames are divided intoWZframes and key frames H264 is used for intracoding anddecoding of key frame meanwhile WZ frame is encoded byWyner-Ziv encoder and the core steps of decoding processare given as follows because of its complexity

Step 1 Extract side information from key frames

Step 2 Simulate the extracted side information by virtualchannel to generate the correlated noise

International Journal of Digital Multimedia Broadcasting 3

H264intraencoder

H264intradecoder

Side information extraction

Virtual channel model

Bit ordering

Minimum rate estimation Soft input

computation

WZ frame

Key frame

Channel encoder

Channeldecoder

Decoder Succfailure

WZ and key frame splitting

Buffer

Encoder Decoder

GOP

T

T

QReconstQ

minus1 andT

minus1

Figure 2 DISCOVER codec architecture

Measurement matrix construction

Grouping Splitter DPCM-NQ Huffman

PackagingInput video

sequence

Nonkeyframe

Packet

Block CS measuring

Blockpartitioning

Block CS measuring

BlockpartitioningKey

frameGOPi

xK k

K

yKk

yNKk

NK

xNKk

k = 1 K

k = 1 K

Figure 3 CSVC encoder architecture

Step 3 Perform the soft input calculation on the correlatednoise and the transformed side information

Step 4 Verify the soft input calculation result by the infor-mation transmitted from encoder side to judge whether thedecoding is successful or not

In the channel coding process DISCOVER uses LowDensity Parity Check Accumulation (LDPCA) code whichis rate-compatible and is closer to the capacity of all types ofchannels when compared with Turbo code The complexityof overall system will be increased because of the requestof cumulative syndrome sent from decoder side to encodersideTherefore theminimumquantity of syndrome that eachbit plane can transmit is set based on the Wyner-Ziv rate-distortion constraint in DISCOVER decoder side to reducethe number of requests and hence obtain higher compressionefficiency

3 Compressive-Sensing Video Coding

CSVC system follows WZ video coding system which isproposed by Wyner-Ziv et al [4] in the way that it alsodivides video stream into key frames and nonkey framesand two different methods are used to implement encodingand decoding of the two types of frames For key framestraditional video intraframe coding framework is adopted orhigh-rate CS codec is introduced to ensure the high qualityof reconstructed key frames For nonkey frames low-rate CSmeasurement is adopted Side information is first extracted

from high-quality key frames and then is combined withthe measurement vectors for joint reconstruction of nonkeyframes In the early research of Distributed CompressedSensing (DCS) video DCS theory was first proposed in [19]which also demonstrated the possibility of the combination ofdistributed coding and CS Since then domestic and foreignscholars have devoted much attention to the research of DCSvideo Typical examples are DIStributed video Coding UsingCompressed Sampling (DISCUCS) proposed by Prades-Nebot et al [20] DIStributed video COmpression Sensing(DISCOS) proposed byDo et al [21] and improvedDISCOSproposed by Tramel and Fowler [22] On the basis of theabove research we propose the CSVC system with superiorperformance in [14] In this paper we will evaluate H264DISCOVER and CSVC system in terms of rate-distortionperformance and rate-energy-distortion performance Thecodec process of CSVC system is described in detail below

31 Encoder Framework The encoder framework of CSVCsystem is shown in Figure 3 First the input video sequenceis divided into several GOPs and key frames and nonkeyframes are separated in the group Then the key framesand the nonkey frames are divided into 119870 nonoverlappingsubblocks of size 119861 times 119861 pixels and each block is arrangedin raster order as column vector of length 1198612 (1198612 = 119873)Finally the measurement matrix is constructed to calculatethe measurement vector of each block as follows

yK119896 = ΦK sdot xK119896 119896 = 1 119870yNK119896 = ΦNK sdot xNK119896 119896 = 1 119870 (1)

4 International Journal of Digital Multimedia Broadcasting

Regrouping

Reference frame selection

Combination

Intrareconstruction

SI prediction

Residual recovery

Unpackaging

Measurement matrix construction

Reconstructed video sequence Joint reconstruction

Nonkey frame

SI

Packet

Keyframe

GOPi

횽K

횽NK

DPCM-NQminus1 Huffmanminus1

yKk

yNKk

Figure 4 CSVC decoder architecture

where xK119896 and xNK119896 denote the 119896th subblock of the keyframe and the nonkey frame respectively andΦK119896 andΦNK119896are the measurement matrix constructed by the randomHadamard matrix For the key frames the size of themeasurementmatrixΦK119896 is119872Ktimes119873 so themeasurement rateis 119878K = 119872K119873 For nonkey frames the size of measurementmatrix ΦNK119896 is119872NK times 119873 so the measurement rate is 119878NK =119872NK119873

Then the measurement vectors yK and yNK of blocks willbe transmitted to the quantizer to form bit stream CSVCsystem uses the nonuniform quantizer based on DifferentialPulse-Code Modulation (DPCM) DPCM-NQ for shortwhich first computes the residual of measurement valuebetween adjacent subblocks to reduce coding redundancyand then quantifies the residual In consideration of thehigh frequency of small residual value the nonuniformquantization is used to process the residual of each subblockin order to reduce the quantization error Supposing that the119898thmeasurement residual of the 119896th subblock is119889119896(119898) it canbe compressed according to 120583 law as follows

119889comp = 119891 [119889119896 (119898)]= sgn [119889119896 (119898)] sdot log (1 + 120583

1003816100381610038161003816119889119896 (119898) 1198631003816100381610038161003816)log (1 + 120583) (2)

where 119863 is the maximum measurement residual of thecurrent frame sgn[sdot] represents the symbolic function and 120583is 10 In the inverse quantization the estimated value of 119889119896(119898)is calculated using the following decompression formula

119889119896 (119898) = 119891minus1 (119889comp)

= sgn [119889comp] sdot 119863120583 sdot [(1 + 120583)119889comp minus 1]

(3)

After the residual of each block is quantized the quantizeddata of all the blocks are subjected to Huffman coding andare encapsulated into data packets to be sent to the decoderside

32 Decoder Framework The decoder framework of CSVCsystem is shown in Figure 4 After the data packets are

received on the decoder side the measurement vectors yK119896of key frames and yNK119896 of nonkey frames can be obtained byHuffman decoding and inverse quantization For key framesthe intraframe reconstruction model is used as follows

xK = argminx 1003817100381710038171003817yK minusΘKE sdot x10038171003817100381710038172 + 120582 Ψ sdot x1 (4)

where

yK =[[[[[[[[[

yK1yK2

yK119870

]]]]]]]]]

ΘK =[[[[[[[[

ΦK 0ΦK

ΦK

0 ΦK

]]]]]]]]

E sdot x =[[[[[[[[[

xK1xK2

xK119870

]]]]]]]]]

(5)

Ψ is the sparse transform matrix of the video frame x and 120582represents the regularization factorThe reconstructedmodel(4) can be solved by a variety of still image CS reconstructionalgorithms To ensure high-quality recovery of key framesCSVC system uses multihypothesis smoothing Landweberiterative algorithm used in [22] to solve model (4)

For the nonkey frame we firstly obtain the side infor-mation xSI of the current nonkey frame by carrying out theside information prediction of the adjacent reconstructed keyframe and then calculate the residual measurement vector

International Journal of Digital Multimedia Broadcasting 5

between the measurement vector of each block and its sideinformation as follows

yR =[[[[[[[

yR1yR2

yR119870

]]]]]]]

=[[[[[[[

yNK1 minusΦNKxSI1yNK2 minusΦNKxSI2

yNK119870 minusΦNKxSI119870

]]]]]]]

=[[[[[[[

ΦNK sdot (xNK1 minus xSI1)ΦNK sdot (xNK2 minus xSI2)

ΦNK sdot (xNK119870 minus xSI119870)

]]]]]]]

(6)

where

ΘNK =[[[[[[

ΦNK 0ΦNK

d

0 ΦNK

]]]]]]

rNK119894 = xNK119894 minus xSI119894

(7)

So (6) can be transformed into

yR = ΘNKE sdot rNK (8)

where rNK is the residual between nonkey frame xNK and sideinformation xSI According to (8) the residual reconstructionmodel of nonkey frame can be established as follows

rNK = argminr1003817100381710038171003817yR minusΘNKE sdot r10038171003817100381710038172 + 120578 P sdot r1 (9)

where P is the sparse transform matrix of the residualrNK and 120578 denotes the regularization factor The residualreconstruction model (9) is still solved using Landweberiterative algorithm Finally the reconstruction of nonkeyframe can be calculated as follows

xNK = xSI + rNK (10)

The features of CSVC system constructed according tothe above codec process are as follows (1) compared toDISCOVER CSVC system eliminates virtual channel andfeedback channel and thus the difficulty of engineering isreduced (2) since there is no correlation between the CSmeasurement and image content the code rate is determinedonly by the measurement rate which makes it easier forCSVC system to control the code rate (3) each measurementvalue contains all the image information therefore it is easyto implement scalable coding (4) the data security can beenhanced by the random generation of the measurementmatrix The above features endow CSVC system with moreengineering value and make it become a potential newDVC scheme We are more concerned about the comparisonof coding energy consumption between CSVC system andH264 andDISCOVER so in the experiment part the codingenergy consumption of the three systems will be evaluated indetail

4 Experimental Results and Analysis

The performances of H264 DISCOVER and CSVC areevaluated respectively using four standard video sequencesnamed Foreman Bus Mobile and Football in CIF for-mat H264 adopts the standard coding configuration ofJM190 model and implements intramode DISCOVER usesthe default encoding configuration and CSVC adopts theexperimental parameter configuration in [14] The rate-distortion and rate-energy-distortion performances of thethree encoders are compared where rate-distortion reflectsthe relationship between the code rate and the Peak Signal-Noise Ratio (PSNR) while rate-energy-distortion reflects therelationship between the coding time and PSNR Using thesame experimental platform the coding time is proportionalto the energy consumption therefore it can represent thelevel of coding energy consumption The experimental plat-form is MATALB R2012b the computer system is 64-bitWindows 7 operating system with an installation memory of800GB and Intel Core i7-4900 processor whose frequency is360GHz

41 Evaluation on Rate-Distortion Performance Figure 5shows the rate-distortion curves for H264 DISCOVER andCSVC encoders under different test video sequences It canbe seen from Figure 5 that the PSNR values of the wholereconstructed video processed by different encoder alwaysgrow in a positive trend when the code rates increase Onthe whole the encoding effects of H264 and DISCOVER arealways better than CSVC encoder For the test videos Busand Football at the same code rate the video reconstructioneffect of H264 is the best for Foreman and Mobile at thesame code rate and in the specific code rate range thecoding effect of DISCOVER is even better than H264 Itcan be seen that the rate-distortion performance of H264 isoptimal DISCOVER follows and CSVC is the worst and hasa big performance difference from the other two For CSVCthe measurement rate determines the bitrate When themeasurement rate is 005 the bitrate is about 6000 kbitss Ifwe further decrease the measurement rate the bitrate will belower than 6000 kbitss The average PSNR of reconstructedvideo gradually decreases with the measurement rate linearlydecreasing The variation of PSNR curve is smooth and thePSNR value cannot suddenly reduce when the bitrate dropsto below 6000 kbitss

42 Evaluation on Rate-Energy-Distortion Performance Fig-ure 6 shows the rate-energy-distortion curves for H264DISCOVER and CSVC encoders under different test videosequences It can be seen from Figure 6 that for any videounder the same PSNR value the encoding time of CSVC isthe shortest DISCOVER follows and H264 is the longestIn particular the average encoding time of CSVC is onlyabout 3 seconds which means that the energy consumptionof CSVC is much lower than DISCOVER and H264 on thesame recovery level With the PSNR value of reconstructedvideo increasing the encoder time of H264 and DISCOVERgradually increases But the change of encoder time underDISCOVER framework is steeper and H264 framework

6 International Journal of Digital Multimedia Broadcasting

H264DISCOVERCSVC

1000 2000 3000 4000 5000 6000 7000 8000 9000 100000Bitrate (kbitss)

30

35

40

45

50

55

60PS

NR

(dB)

(a) Foreman

2000 4000 6000 8000 10000 120000Bitrate (kbitss)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(b) Bus

H264DISCOVERCSVC

2000 4000 6000 8000 10000 120000Bitrate (kbitss)

25

30

35

40

45

50

55

60

PSN

R (d

B)

(c) Football

25

30

35

40

45

50

55

PSN

R (d

B)

5000 10000 150000Bitrate (kbitss)

H264DISCOVERCSVC

(d) Mobile

Figure 5 Comparison of rate-distortion performance of H264 DISCOVER and CSVC encoder under different test video sequences

more gentle which shows that H264 has a high dependencyon energy consumption with its promotion of performanceand DISCOVER also needs a certain amount of energy inputThe computational complexity of CS measuring determinesthe energy consumption at encoder Suppose119872 denotes thenumber of CSmeasurements for a video frame and 119871 denotesthe total number of pixels in a video frame The computa-tional complexity of CS measuring is 119874(119872119871) Because119872 isfar below 119871 the variation of energy consumption is very smallwhen changing 119872 However 119872 is an important factor forthe reconstruction quality of video frameThe reconstructionquality can be improved effectively with small increments of119872 Therefore the slope of the rate-energy-distortion curveis almost vertical indicating that the small investment ofenergy consumption can get the significant improvement ofreconstruction quality It can be seen that the rate-energy-distortion performance of CSVC is optimal DISCOVERfollows H264 is the worst Among the three the energyconsumption of CSVC is approximately invariant regardlessof reconstruction quality while the reconstruction quality ofH264 andDISCOVERhas a great correlationwith the energyconsumption of coding

5 Conclusions

This paper has conducted an experiment-driven analysis ofrate-distortion and rate-energy-distortion performances ofCSVC algorithm and compares them with that of H264 andDISCOVER The rate-distortion and rate-energy-distortionperformances of the three systems are evaluated under thesame experimental environment Experiment results showthat the rate-distortion performance of CSVC has a largeperformance difference from H264 and DISCOVER but itsrate-energy-distortion performance has a greater advantagethat is the rapid improvement of its reconstruction qualitydoes not depend on coding energy input Therefore onthe premise that communication bandwidth is effectivelyimproved CSVC can be used as a candidate for futurewireless video communication because of its characteristicswhich provides wireless video terminals limited by energyconsumption and computing power with more possibili-ties

At present the rate-distortion performance of CSVC isstill not ideal and there is still some way to go before weput CSVC into practical use Efforts should be made in the

International Journal of Digital Multimedia Broadcasting 7

30

35

40

45

50

55

60PS

NR

(dB)

50 100 150 200 250 300 3500Encoder time (s)

H264DISCOVERCSVC

(a) Foreman

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(b) Bus

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

60

PSN

R (d

B)

H264DISCOVERCSVC

(c) Football

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(d) Mobile

Figure 6 Comparison of rate-energy-distortion performance ofH264 DISCOVER andCSVC encoder under different test video sequences

following areas to improve the rate-distortion performanceof CSVC

(1) Side Information Estimation Rate-distortion performanceof CSVC is greatly related to the accuracy of side informationestimation which means that high-quality side informationimmensely reduces the required supply of bit load fromencoder side Therefore finding the appropriate motion esti-mation algorithm to obtain more accurate side informationand realizing the optimal reconstruction of decoding willbecome the key to improving the rate-distortion performanceof CSVC

(2) A Priori Structural Feature Modeling of Video FramesImages of the same type often have similar structuralinformation Therefore the reconstructed model can beconstructed by evaluating the structural information of thedecoded video frames and extracting the prior knowledgewhich can reduce code rate and improve the reconstructionquality For example the statistical correlation structurecan be used in the image transformation coefficient and atree structure can be adopted for the wavelet coefficientHowever due to the complexity and uncertainty of natural

images further study should be made on how to use a prioriknowledge to construct a suitable model

(3) Quantization Measurement Uniform quantization is themajor method adopted to quantify the CS measurementcurrently But the traditional entropy coding method isnot ideal for compressing the uniform quantization valuesbecause of the statistical independence between the uniformquantization values Then how to express the CS value withthe least number of bits with the constraint of information-theoretic rate-distortion coding theorem is one of the keytopics of the following research Therefore it is necessary topropose a new nonuniform quantization method to establishstatistical correlation between quantization values and herebyto design a new entropy coding method matching thestatistical correlation

The above-mentioned further researches are employedat decoder to improve the rate-distortion performance ofCSVC but the CS measuring at encoder guarantees theadvantage of CSVC in rate-energy-distortion performanceTherefore the rate-energy-distortion performance cannot beaffected while improving the rate-distortion performance ofCSVC

8 International Journal of Digital Multimedia Broadcasting

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work was supported in part by the National NaturalScience Foundation of China under Grant no 61501393in part by the Key Scientific Research Project of Collegesand Universities in Henan Province of China under Grant16A520069 in part by Youth Sustentation Fund of XinyangNormal University under Grant no 2015-QN-043 and inpart by Scientific Research Foundation of Graduate School ofXinyang Normal University under Grant no 2016KYJJ10

References

[1] J Ostermann J Bormans P List et al ldquoVideo coding withH264AVC tools performance and complexityrdquo IEEE Circuitsand Systems Magazine vol 4 no 1 pp 7ndash28 2004

[2] B Girod A M Aaron S Rane and D Rebollo-MonederoldquoDistributed video codingrdquo Proceedings of the IEEE vol 93 no1 pp 71ndash83 2005

[3] A DWyner and J Ziv ldquoThe rate-distortion function for sourcecoding with side information at the decoderrdquo IEEE Transactionson Information Theory vol 22 no 1 pp 1ndash10 1976

[4] A Aaron S Rane R Zhang and B Girod ldquoWyner-Ziv codingfor video applications to compression and error resiliencerdquo inProceedings of the Data Compression Conference (DCC rsquo03) pp93ndash102 Snowbird Utah USA March 2003

[5] R Puri A Majumdar and K Ramchandran ldquoPRISM a videocoding paradigm with motion estimation at the decoderrdquo IEEETransactions on Image Processing vol 16 no 10 pp 2436ndash24482007

[6] Q Xu and Z Xiong ldquoLayered Wyner-Ziv video codingrdquo inProceedings of the Visual Communications and Image Processing2004 pp 83ndash91 IEEE San Jose Calif USA January 2004

[7] W Liu L Dong and W Zeng ldquoMotion refinement basedprogressive side-information estimation for Wyner-Ziv videocodingrdquo IEEE Transactions on Circuits amp Systems for VideoTechnology vol 20 no 12 pp 1863ndash1875 2010

[8] X Artigas J Ascenso M Dalai S Klomp D Kubasov and MOuaret ldquoThe discover codec architecture techniques and eval-uationrdquo in Proceedings of the 26th Picture Coding Symposium(PCS rsquo07) pp 1103ndash1120 Lisbon Portugal November 2007

[9] J Slowack J Skorupa N Deligiannis P Lambert AMunteanuand R van de Walle ldquoDistributed video coding with feedbackchannel constraintsrdquo IEEE Transactions on Circuits amp Systemsfor Video Technology vol 22 no 7 pp 1014ndash1026 2012

[10] C Brites andF Pereira ldquoCorrelationnoisemodeling for efficientpixel and transform domain Wyner-Ziv video codingrdquo IEEETransactions on Circuits and Systems for Video Technology vol18 no 9 pp 1177ndash1190 2008

[11] E J Candes J Romberg and T Tao ldquoRobust uncertaintyprinciples exact signal reconstruction from highly incompletefrequency informationrdquo IEEE Transactions on InformationThe-ory vol 52 no 2 pp 489ndash509 2006

[12] D L Donoho ldquoCompressed sensingrdquo IEEE Transactions onInformation Theory vol 52 no 4 pp 1289ndash1306 2006

[13] E J Candes and M B Wakin ldquoAn introduction to compressivesampling a sensingsampling paradigm that goes against thecommon knowledge in data acquisitionrdquo IEEE Signal ProcessingMagazine vol 25 no 2 pp 21ndash30 2008

[14] R Li H Liu R Xue and Y Li ldquoCompressive-sensing-basedvideo codec by autoregressive prediction and adaptive residualrecoveryrdquo International Journal of Distributed Sensor Networksvol 2015 Article ID 562840 19 pages 2015

[15] Y Liu X Zhu L Zhang and S H Cho ldquoDistributed com-pressed video sensing in camera sensor networksrdquo InternationalJournal of Distributed Sensor Networks vol 2012 Article ID352167 10 pages 2012

[16] C Di Laura D Pajuelo andG Kemper ldquoA novel steganographytechnique for SDTV-H264AVC encoded videordquo InternationalJournal of Digital Multimedia Broadcasting vol 2016 Article ID6950592 9 pages 2016

[17] Y Lahbabi E H Ibn Elhaj and A Hammouch ldquoAdaptivestreaming of scalable videos over P2PTVrdquo International Journalof Digital Multimedia Broadcasting vol 2015 Article ID 28309710 pages 2015

[18] S Pudlewski and T Melodia ldquoCompressive video streamingdesign and rate-energy-distortion analysisrdquo IEEE Transactionson Multimedia vol 15 no 8 pp 2072ndash2086 2013

[19] D Baron M B Wakin M F Duarte S Sarvotham and R GBaraniuk ldquoDistributed compressive sensingrdquo httparxivorgabs09013403

[20] J Prades-Nebot Y Ma and T Huang ldquoDistributed videocoding using compressive samplingrdquo in Proceedings of thePicture Coding Symposium (PCS rsquo09) pp 1ndash4 Chicago Ill USAMay 2009

[21] T T Do Y Chen D T Nguyen N Nguyen L Gan and T DTran ldquoDistributed compressed video sensingrdquo in Proceedingsof the 16th IEEE International Conference on Image Processing(ICIP rsquo09) pp 1393ndash1396 IEEE Cairo Egypt November 2009

[22] E W Tramel and J E Fowler ldquoVideo compressed sensingwith multihypothesisrdquo in Proceedings of the Data CompressionConference (DCC rsquo11) pp 193ndash202 Snowbird Utah USAMarch 2011

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpswwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 2: Rate-Distortion and Rate-Energy-Distortion Evaluations of ...

2 International Journal of Digital Multimedia Broadcasting

MC

Intrapredictionselection

Intraprediction

ME

Inter

Intra NAL

Entropy coding

Reordering

Filtering

Current block Fn

Reference block F㰀

nminus1

Reconstructed block F㰀

n

Dn

T Q

X

P

D㰀n

Tminus1 Q

minus1

minus

+

uF㰀n

Figure 1 H264 codec architecture

H264 and DISCOVER It will help to clarify the upperlimit of performance improvement of CSVC by obtaining theresult of the performance difference between the three videocoding schemes The rate-distortion performance cannotshow the relationship between coding energy consumptionand video reconstruction quality Therefore rate-energy-distortion performance [18] also needs an objective eval-uation of CSVC H264 and DISCOVER This paper firstsummarizes the basic framework and technical details ofH264 and DISCOVER Next a typical algorithm of CSVCwill be described in detail Finally on the basis of the theoriesexposition rate-distortion performance and rate-energy-distortion performance of the three video coding schemesare evaluated and the performance difference between thethree video coding schemes is then fully discussed Theexperimental results show that under the test of CIF videosrespectively named Bus Football Foreman and Mobile interms of the rate-distortion performance H264 is optimaland DISCOVER follows while CSVC is the worst and hasa large performance difference from the other two but interms of rate-energy-distortion performance CSVC is opti-mal DISCOVER follows and H264 is the worst The resultsalso reveal a fact that the energy consumption of CSVC isapproximately the same regardless of reconstruction qualitywhile for H264 and DISCOVER there is a close correlationbetween recovery quality and the energy consumption ofcoding As a result it can be concluded that when low energyconsumption is demanded CSVC program can give full playto its advantages but its rate-distortion performance stillneeds improvement

2 Typical Video Coding Schemes

21 H264 H264 system is divided into two levels in func-tion Video Coding Layer (VCL) and Network AbstractionLayer (NAL) Its codec framework is shown in Figure 1 In thecoding process there are two options to predict the currentimage block F119899 interprediction and intraframe predictionWhen interprediction is adopted the motion vector of cur-rent block F119899 is obtained by the motion estimation accordingto the reference block and then the predicted frame Pcould be obtained by the motion compensation method

when intraprediction is used the predicted block in thecurrent frame is the weighted average of the selected adjacentdecoded blocks of the current block After the predictiveframe is determined the main steps of codec process are asfollows

Step 1 Calculate the residual D119899 between the current blockF119899 and the predicted value P

Step 2 Obtain the quantized coefficient 119883 by transformingand quantizingD119899

Step 3 Form the bit stream by reordering and entropy codingof quantized coefficient119883Step 4 Transmit the bit stream to the decoder side throughNAL at the same time use part of the bit stream which couldbe decoded on the encoder side as reference frame

The core technology of H264 is mainly reflected on itsimprovement in the interframe and intraframe predictivecoding for example using 4 times 4 integer discrete cosinetransform technology instead of the former 8 times 8 discretecosine transform technology to avoid mismatches in inversetransformation

22 DISCOVER The codec framework of DISCOVER sys-tem is shown in Figure 2 The encoder side carries on videoprocessing in Group of Pictures (GOP) whose length isdetermined by the specific situation The length of GOP willbe increased to reduce time redundancy when the imagecontains a small amount of motion on the other hand thelength can be shortened accordingly if there is a large amountof motion For each group video frames are divided intoWZframes and key frames H264 is used for intracoding anddecoding of key frame meanwhile WZ frame is encoded byWyner-Ziv encoder and the core steps of decoding processare given as follows because of its complexity

Step 1 Extract side information from key frames

Step 2 Simulate the extracted side information by virtualchannel to generate the correlated noise

International Journal of Digital Multimedia Broadcasting 3

H264intraencoder

H264intradecoder

Side information extraction

Virtual channel model

Bit ordering

Minimum rate estimation Soft input

computation

WZ frame

Key frame

Channel encoder

Channeldecoder

Decoder Succfailure

WZ and key frame splitting

Buffer

Encoder Decoder

GOP

T

T

QReconstQ

minus1 andT

minus1

Figure 2 DISCOVER codec architecture

Measurement matrix construction

Grouping Splitter DPCM-NQ Huffman

PackagingInput video

sequence

Nonkeyframe

Packet

Block CS measuring

Blockpartitioning

Block CS measuring

BlockpartitioningKey

frameGOPi

xK k

K

yKk

yNKk

NK

xNKk

k = 1 K

k = 1 K

Figure 3 CSVC encoder architecture

Step 3 Perform the soft input calculation on the correlatednoise and the transformed side information

Step 4 Verify the soft input calculation result by the infor-mation transmitted from encoder side to judge whether thedecoding is successful or not

In the channel coding process DISCOVER uses LowDensity Parity Check Accumulation (LDPCA) code whichis rate-compatible and is closer to the capacity of all types ofchannels when compared with Turbo code The complexityof overall system will be increased because of the requestof cumulative syndrome sent from decoder side to encodersideTherefore theminimumquantity of syndrome that eachbit plane can transmit is set based on the Wyner-Ziv rate-distortion constraint in DISCOVER decoder side to reducethe number of requests and hence obtain higher compressionefficiency

3 Compressive-Sensing Video Coding

CSVC system follows WZ video coding system which isproposed by Wyner-Ziv et al [4] in the way that it alsodivides video stream into key frames and nonkey framesand two different methods are used to implement encodingand decoding of the two types of frames For key framestraditional video intraframe coding framework is adopted orhigh-rate CS codec is introduced to ensure the high qualityof reconstructed key frames For nonkey frames low-rate CSmeasurement is adopted Side information is first extracted

from high-quality key frames and then is combined withthe measurement vectors for joint reconstruction of nonkeyframes In the early research of Distributed CompressedSensing (DCS) video DCS theory was first proposed in [19]which also demonstrated the possibility of the combination ofdistributed coding and CS Since then domestic and foreignscholars have devoted much attention to the research of DCSvideo Typical examples are DIStributed video Coding UsingCompressed Sampling (DISCUCS) proposed by Prades-Nebot et al [20] DIStributed video COmpression Sensing(DISCOS) proposed byDo et al [21] and improvedDISCOSproposed by Tramel and Fowler [22] On the basis of theabove research we propose the CSVC system with superiorperformance in [14] In this paper we will evaluate H264DISCOVER and CSVC system in terms of rate-distortionperformance and rate-energy-distortion performance Thecodec process of CSVC system is described in detail below

31 Encoder Framework The encoder framework of CSVCsystem is shown in Figure 3 First the input video sequenceis divided into several GOPs and key frames and nonkeyframes are separated in the group Then the key framesand the nonkey frames are divided into 119870 nonoverlappingsubblocks of size 119861 times 119861 pixels and each block is arrangedin raster order as column vector of length 1198612 (1198612 = 119873)Finally the measurement matrix is constructed to calculatethe measurement vector of each block as follows

yK119896 = ΦK sdot xK119896 119896 = 1 119870yNK119896 = ΦNK sdot xNK119896 119896 = 1 119870 (1)

4 International Journal of Digital Multimedia Broadcasting

Regrouping

Reference frame selection

Combination

Intrareconstruction

SI prediction

Residual recovery

Unpackaging

Measurement matrix construction

Reconstructed video sequence Joint reconstruction

Nonkey frame

SI

Packet

Keyframe

GOPi

횽K

횽NK

DPCM-NQminus1 Huffmanminus1

yKk

yNKk

Figure 4 CSVC decoder architecture

where xK119896 and xNK119896 denote the 119896th subblock of the keyframe and the nonkey frame respectively andΦK119896 andΦNK119896are the measurement matrix constructed by the randomHadamard matrix For the key frames the size of themeasurementmatrixΦK119896 is119872Ktimes119873 so themeasurement rateis 119878K = 119872K119873 For nonkey frames the size of measurementmatrix ΦNK119896 is119872NK times 119873 so the measurement rate is 119878NK =119872NK119873

Then the measurement vectors yK and yNK of blocks willbe transmitted to the quantizer to form bit stream CSVCsystem uses the nonuniform quantizer based on DifferentialPulse-Code Modulation (DPCM) DPCM-NQ for shortwhich first computes the residual of measurement valuebetween adjacent subblocks to reduce coding redundancyand then quantifies the residual In consideration of thehigh frequency of small residual value the nonuniformquantization is used to process the residual of each subblockin order to reduce the quantization error Supposing that the119898thmeasurement residual of the 119896th subblock is119889119896(119898) it canbe compressed according to 120583 law as follows

119889comp = 119891 [119889119896 (119898)]= sgn [119889119896 (119898)] sdot log (1 + 120583

1003816100381610038161003816119889119896 (119898) 1198631003816100381610038161003816)log (1 + 120583) (2)

where 119863 is the maximum measurement residual of thecurrent frame sgn[sdot] represents the symbolic function and 120583is 10 In the inverse quantization the estimated value of 119889119896(119898)is calculated using the following decompression formula

119889119896 (119898) = 119891minus1 (119889comp)

= sgn [119889comp] sdot 119863120583 sdot [(1 + 120583)119889comp minus 1]

(3)

After the residual of each block is quantized the quantizeddata of all the blocks are subjected to Huffman coding andare encapsulated into data packets to be sent to the decoderside

32 Decoder Framework The decoder framework of CSVCsystem is shown in Figure 4 After the data packets are

received on the decoder side the measurement vectors yK119896of key frames and yNK119896 of nonkey frames can be obtained byHuffman decoding and inverse quantization For key framesthe intraframe reconstruction model is used as follows

xK = argminx 1003817100381710038171003817yK minusΘKE sdot x10038171003817100381710038172 + 120582 Ψ sdot x1 (4)

where

yK =[[[[[[[[[

yK1yK2

yK119870

]]]]]]]]]

ΘK =[[[[[[[[

ΦK 0ΦK

ΦK

0 ΦK

]]]]]]]]

E sdot x =[[[[[[[[[

xK1xK2

xK119870

]]]]]]]]]

(5)

Ψ is the sparse transform matrix of the video frame x and 120582represents the regularization factorThe reconstructedmodel(4) can be solved by a variety of still image CS reconstructionalgorithms To ensure high-quality recovery of key framesCSVC system uses multihypothesis smoothing Landweberiterative algorithm used in [22] to solve model (4)

For the nonkey frame we firstly obtain the side infor-mation xSI of the current nonkey frame by carrying out theside information prediction of the adjacent reconstructed keyframe and then calculate the residual measurement vector

International Journal of Digital Multimedia Broadcasting 5

between the measurement vector of each block and its sideinformation as follows

yR =[[[[[[[

yR1yR2

yR119870

]]]]]]]

=[[[[[[[

yNK1 minusΦNKxSI1yNK2 minusΦNKxSI2

yNK119870 minusΦNKxSI119870

]]]]]]]

=[[[[[[[

ΦNK sdot (xNK1 minus xSI1)ΦNK sdot (xNK2 minus xSI2)

ΦNK sdot (xNK119870 minus xSI119870)

]]]]]]]

(6)

where

ΘNK =[[[[[[

ΦNK 0ΦNK

d

0 ΦNK

]]]]]]

rNK119894 = xNK119894 minus xSI119894

(7)

So (6) can be transformed into

yR = ΘNKE sdot rNK (8)

where rNK is the residual between nonkey frame xNK and sideinformation xSI According to (8) the residual reconstructionmodel of nonkey frame can be established as follows

rNK = argminr1003817100381710038171003817yR minusΘNKE sdot r10038171003817100381710038172 + 120578 P sdot r1 (9)

where P is the sparse transform matrix of the residualrNK and 120578 denotes the regularization factor The residualreconstruction model (9) is still solved using Landweberiterative algorithm Finally the reconstruction of nonkeyframe can be calculated as follows

xNK = xSI + rNK (10)

The features of CSVC system constructed according tothe above codec process are as follows (1) compared toDISCOVER CSVC system eliminates virtual channel andfeedback channel and thus the difficulty of engineering isreduced (2) since there is no correlation between the CSmeasurement and image content the code rate is determinedonly by the measurement rate which makes it easier forCSVC system to control the code rate (3) each measurementvalue contains all the image information therefore it is easyto implement scalable coding (4) the data security can beenhanced by the random generation of the measurementmatrix The above features endow CSVC system with moreengineering value and make it become a potential newDVC scheme We are more concerned about the comparisonof coding energy consumption between CSVC system andH264 andDISCOVER so in the experiment part the codingenergy consumption of the three systems will be evaluated indetail

4 Experimental Results and Analysis

The performances of H264 DISCOVER and CSVC areevaluated respectively using four standard video sequencesnamed Foreman Bus Mobile and Football in CIF for-mat H264 adopts the standard coding configuration ofJM190 model and implements intramode DISCOVER usesthe default encoding configuration and CSVC adopts theexperimental parameter configuration in [14] The rate-distortion and rate-energy-distortion performances of thethree encoders are compared where rate-distortion reflectsthe relationship between the code rate and the Peak Signal-Noise Ratio (PSNR) while rate-energy-distortion reflects therelationship between the coding time and PSNR Using thesame experimental platform the coding time is proportionalto the energy consumption therefore it can represent thelevel of coding energy consumption The experimental plat-form is MATALB R2012b the computer system is 64-bitWindows 7 operating system with an installation memory of800GB and Intel Core i7-4900 processor whose frequency is360GHz

41 Evaluation on Rate-Distortion Performance Figure 5shows the rate-distortion curves for H264 DISCOVER andCSVC encoders under different test video sequences It canbe seen from Figure 5 that the PSNR values of the wholereconstructed video processed by different encoder alwaysgrow in a positive trend when the code rates increase Onthe whole the encoding effects of H264 and DISCOVER arealways better than CSVC encoder For the test videos Busand Football at the same code rate the video reconstructioneffect of H264 is the best for Foreman and Mobile at thesame code rate and in the specific code rate range thecoding effect of DISCOVER is even better than H264 Itcan be seen that the rate-distortion performance of H264 isoptimal DISCOVER follows and CSVC is the worst and hasa big performance difference from the other two For CSVCthe measurement rate determines the bitrate When themeasurement rate is 005 the bitrate is about 6000 kbitss Ifwe further decrease the measurement rate the bitrate will belower than 6000 kbitss The average PSNR of reconstructedvideo gradually decreases with the measurement rate linearlydecreasing The variation of PSNR curve is smooth and thePSNR value cannot suddenly reduce when the bitrate dropsto below 6000 kbitss

42 Evaluation on Rate-Energy-Distortion Performance Fig-ure 6 shows the rate-energy-distortion curves for H264DISCOVER and CSVC encoders under different test videosequences It can be seen from Figure 6 that for any videounder the same PSNR value the encoding time of CSVC isthe shortest DISCOVER follows and H264 is the longestIn particular the average encoding time of CSVC is onlyabout 3 seconds which means that the energy consumptionof CSVC is much lower than DISCOVER and H264 on thesame recovery level With the PSNR value of reconstructedvideo increasing the encoder time of H264 and DISCOVERgradually increases But the change of encoder time underDISCOVER framework is steeper and H264 framework

6 International Journal of Digital Multimedia Broadcasting

H264DISCOVERCSVC

1000 2000 3000 4000 5000 6000 7000 8000 9000 100000Bitrate (kbitss)

30

35

40

45

50

55

60PS

NR

(dB)

(a) Foreman

2000 4000 6000 8000 10000 120000Bitrate (kbitss)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(b) Bus

H264DISCOVERCSVC

2000 4000 6000 8000 10000 120000Bitrate (kbitss)

25

30

35

40

45

50

55

60

PSN

R (d

B)

(c) Football

25

30

35

40

45

50

55

PSN

R (d

B)

5000 10000 150000Bitrate (kbitss)

H264DISCOVERCSVC

(d) Mobile

Figure 5 Comparison of rate-distortion performance of H264 DISCOVER and CSVC encoder under different test video sequences

more gentle which shows that H264 has a high dependencyon energy consumption with its promotion of performanceand DISCOVER also needs a certain amount of energy inputThe computational complexity of CS measuring determinesthe energy consumption at encoder Suppose119872 denotes thenumber of CSmeasurements for a video frame and 119871 denotesthe total number of pixels in a video frame The computa-tional complexity of CS measuring is 119874(119872119871) Because119872 isfar below 119871 the variation of energy consumption is very smallwhen changing 119872 However 119872 is an important factor forthe reconstruction quality of video frameThe reconstructionquality can be improved effectively with small increments of119872 Therefore the slope of the rate-energy-distortion curveis almost vertical indicating that the small investment ofenergy consumption can get the significant improvement ofreconstruction quality It can be seen that the rate-energy-distortion performance of CSVC is optimal DISCOVERfollows H264 is the worst Among the three the energyconsumption of CSVC is approximately invariant regardlessof reconstruction quality while the reconstruction quality ofH264 andDISCOVERhas a great correlationwith the energyconsumption of coding

5 Conclusions

This paper has conducted an experiment-driven analysis ofrate-distortion and rate-energy-distortion performances ofCSVC algorithm and compares them with that of H264 andDISCOVER The rate-distortion and rate-energy-distortionperformances of the three systems are evaluated under thesame experimental environment Experiment results showthat the rate-distortion performance of CSVC has a largeperformance difference from H264 and DISCOVER but itsrate-energy-distortion performance has a greater advantagethat is the rapid improvement of its reconstruction qualitydoes not depend on coding energy input Therefore onthe premise that communication bandwidth is effectivelyimproved CSVC can be used as a candidate for futurewireless video communication because of its characteristicswhich provides wireless video terminals limited by energyconsumption and computing power with more possibili-ties

At present the rate-distortion performance of CSVC isstill not ideal and there is still some way to go before weput CSVC into practical use Efforts should be made in the

International Journal of Digital Multimedia Broadcasting 7

30

35

40

45

50

55

60PS

NR

(dB)

50 100 150 200 250 300 3500Encoder time (s)

H264DISCOVERCSVC

(a) Foreman

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(b) Bus

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

60

PSN

R (d

B)

H264DISCOVERCSVC

(c) Football

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(d) Mobile

Figure 6 Comparison of rate-energy-distortion performance ofH264 DISCOVER andCSVC encoder under different test video sequences

following areas to improve the rate-distortion performanceof CSVC

(1) Side Information Estimation Rate-distortion performanceof CSVC is greatly related to the accuracy of side informationestimation which means that high-quality side informationimmensely reduces the required supply of bit load fromencoder side Therefore finding the appropriate motion esti-mation algorithm to obtain more accurate side informationand realizing the optimal reconstruction of decoding willbecome the key to improving the rate-distortion performanceof CSVC

(2) A Priori Structural Feature Modeling of Video FramesImages of the same type often have similar structuralinformation Therefore the reconstructed model can beconstructed by evaluating the structural information of thedecoded video frames and extracting the prior knowledgewhich can reduce code rate and improve the reconstructionquality For example the statistical correlation structurecan be used in the image transformation coefficient and atree structure can be adopted for the wavelet coefficientHowever due to the complexity and uncertainty of natural

images further study should be made on how to use a prioriknowledge to construct a suitable model

(3) Quantization Measurement Uniform quantization is themajor method adopted to quantify the CS measurementcurrently But the traditional entropy coding method isnot ideal for compressing the uniform quantization valuesbecause of the statistical independence between the uniformquantization values Then how to express the CS value withthe least number of bits with the constraint of information-theoretic rate-distortion coding theorem is one of the keytopics of the following research Therefore it is necessary topropose a new nonuniform quantization method to establishstatistical correlation between quantization values and herebyto design a new entropy coding method matching thestatistical correlation

The above-mentioned further researches are employedat decoder to improve the rate-distortion performance ofCSVC but the CS measuring at encoder guarantees theadvantage of CSVC in rate-energy-distortion performanceTherefore the rate-energy-distortion performance cannot beaffected while improving the rate-distortion performance ofCSVC

8 International Journal of Digital Multimedia Broadcasting

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work was supported in part by the National NaturalScience Foundation of China under Grant no 61501393in part by the Key Scientific Research Project of Collegesand Universities in Henan Province of China under Grant16A520069 in part by Youth Sustentation Fund of XinyangNormal University under Grant no 2015-QN-043 and inpart by Scientific Research Foundation of Graduate School ofXinyang Normal University under Grant no 2016KYJJ10

References

[1] J Ostermann J Bormans P List et al ldquoVideo coding withH264AVC tools performance and complexityrdquo IEEE Circuitsand Systems Magazine vol 4 no 1 pp 7ndash28 2004

[2] B Girod A M Aaron S Rane and D Rebollo-MonederoldquoDistributed video codingrdquo Proceedings of the IEEE vol 93 no1 pp 71ndash83 2005

[3] A DWyner and J Ziv ldquoThe rate-distortion function for sourcecoding with side information at the decoderrdquo IEEE Transactionson Information Theory vol 22 no 1 pp 1ndash10 1976

[4] A Aaron S Rane R Zhang and B Girod ldquoWyner-Ziv codingfor video applications to compression and error resiliencerdquo inProceedings of the Data Compression Conference (DCC rsquo03) pp93ndash102 Snowbird Utah USA March 2003

[5] R Puri A Majumdar and K Ramchandran ldquoPRISM a videocoding paradigm with motion estimation at the decoderrdquo IEEETransactions on Image Processing vol 16 no 10 pp 2436ndash24482007

[6] Q Xu and Z Xiong ldquoLayered Wyner-Ziv video codingrdquo inProceedings of the Visual Communications and Image Processing2004 pp 83ndash91 IEEE San Jose Calif USA January 2004

[7] W Liu L Dong and W Zeng ldquoMotion refinement basedprogressive side-information estimation for Wyner-Ziv videocodingrdquo IEEE Transactions on Circuits amp Systems for VideoTechnology vol 20 no 12 pp 1863ndash1875 2010

[8] X Artigas J Ascenso M Dalai S Klomp D Kubasov and MOuaret ldquoThe discover codec architecture techniques and eval-uationrdquo in Proceedings of the 26th Picture Coding Symposium(PCS rsquo07) pp 1103ndash1120 Lisbon Portugal November 2007

[9] J Slowack J Skorupa N Deligiannis P Lambert AMunteanuand R van de Walle ldquoDistributed video coding with feedbackchannel constraintsrdquo IEEE Transactions on Circuits amp Systemsfor Video Technology vol 22 no 7 pp 1014ndash1026 2012

[10] C Brites andF Pereira ldquoCorrelationnoisemodeling for efficientpixel and transform domain Wyner-Ziv video codingrdquo IEEETransactions on Circuits and Systems for Video Technology vol18 no 9 pp 1177ndash1190 2008

[11] E J Candes J Romberg and T Tao ldquoRobust uncertaintyprinciples exact signal reconstruction from highly incompletefrequency informationrdquo IEEE Transactions on InformationThe-ory vol 52 no 2 pp 489ndash509 2006

[12] D L Donoho ldquoCompressed sensingrdquo IEEE Transactions onInformation Theory vol 52 no 4 pp 1289ndash1306 2006

[13] E J Candes and M B Wakin ldquoAn introduction to compressivesampling a sensingsampling paradigm that goes against thecommon knowledge in data acquisitionrdquo IEEE Signal ProcessingMagazine vol 25 no 2 pp 21ndash30 2008

[14] R Li H Liu R Xue and Y Li ldquoCompressive-sensing-basedvideo codec by autoregressive prediction and adaptive residualrecoveryrdquo International Journal of Distributed Sensor Networksvol 2015 Article ID 562840 19 pages 2015

[15] Y Liu X Zhu L Zhang and S H Cho ldquoDistributed com-pressed video sensing in camera sensor networksrdquo InternationalJournal of Distributed Sensor Networks vol 2012 Article ID352167 10 pages 2012

[16] C Di Laura D Pajuelo andG Kemper ldquoA novel steganographytechnique for SDTV-H264AVC encoded videordquo InternationalJournal of Digital Multimedia Broadcasting vol 2016 Article ID6950592 9 pages 2016

[17] Y Lahbabi E H Ibn Elhaj and A Hammouch ldquoAdaptivestreaming of scalable videos over P2PTVrdquo International Journalof Digital Multimedia Broadcasting vol 2015 Article ID 28309710 pages 2015

[18] S Pudlewski and T Melodia ldquoCompressive video streamingdesign and rate-energy-distortion analysisrdquo IEEE Transactionson Multimedia vol 15 no 8 pp 2072ndash2086 2013

[19] D Baron M B Wakin M F Duarte S Sarvotham and R GBaraniuk ldquoDistributed compressive sensingrdquo httparxivorgabs09013403

[20] J Prades-Nebot Y Ma and T Huang ldquoDistributed videocoding using compressive samplingrdquo in Proceedings of thePicture Coding Symposium (PCS rsquo09) pp 1ndash4 Chicago Ill USAMay 2009

[21] T T Do Y Chen D T Nguyen N Nguyen L Gan and T DTran ldquoDistributed compressed video sensingrdquo in Proceedingsof the 16th IEEE International Conference on Image Processing(ICIP rsquo09) pp 1393ndash1396 IEEE Cairo Egypt November 2009

[22] E W Tramel and J E Fowler ldquoVideo compressed sensingwith multihypothesisrdquo in Proceedings of the Data CompressionConference (DCC rsquo11) pp 193ndash202 Snowbird Utah USAMarch 2011

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpswwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 3: Rate-Distortion and Rate-Energy-Distortion Evaluations of ...

International Journal of Digital Multimedia Broadcasting 3

H264intraencoder

H264intradecoder

Side information extraction

Virtual channel model

Bit ordering

Minimum rate estimation Soft input

computation

WZ frame

Key frame

Channel encoder

Channeldecoder

Decoder Succfailure

WZ and key frame splitting

Buffer

Encoder Decoder

GOP

T

T

QReconstQ

minus1 andT

minus1

Figure 2 DISCOVER codec architecture

Measurement matrix construction

Grouping Splitter DPCM-NQ Huffman

PackagingInput video

sequence

Nonkeyframe

Packet

Block CS measuring

Blockpartitioning

Block CS measuring

BlockpartitioningKey

frameGOPi

xK k

K

yKk

yNKk

NK

xNKk

k = 1 K

k = 1 K

Figure 3 CSVC encoder architecture

Step 3 Perform the soft input calculation on the correlatednoise and the transformed side information

Step 4 Verify the soft input calculation result by the infor-mation transmitted from encoder side to judge whether thedecoding is successful or not

In the channel coding process DISCOVER uses LowDensity Parity Check Accumulation (LDPCA) code whichis rate-compatible and is closer to the capacity of all types ofchannels when compared with Turbo code The complexityof overall system will be increased because of the requestof cumulative syndrome sent from decoder side to encodersideTherefore theminimumquantity of syndrome that eachbit plane can transmit is set based on the Wyner-Ziv rate-distortion constraint in DISCOVER decoder side to reducethe number of requests and hence obtain higher compressionefficiency

3 Compressive-Sensing Video Coding

CSVC system follows WZ video coding system which isproposed by Wyner-Ziv et al [4] in the way that it alsodivides video stream into key frames and nonkey framesand two different methods are used to implement encodingand decoding of the two types of frames For key framestraditional video intraframe coding framework is adopted orhigh-rate CS codec is introduced to ensure the high qualityof reconstructed key frames For nonkey frames low-rate CSmeasurement is adopted Side information is first extracted

from high-quality key frames and then is combined withthe measurement vectors for joint reconstruction of nonkeyframes In the early research of Distributed CompressedSensing (DCS) video DCS theory was first proposed in [19]which also demonstrated the possibility of the combination ofdistributed coding and CS Since then domestic and foreignscholars have devoted much attention to the research of DCSvideo Typical examples are DIStributed video Coding UsingCompressed Sampling (DISCUCS) proposed by Prades-Nebot et al [20] DIStributed video COmpression Sensing(DISCOS) proposed byDo et al [21] and improvedDISCOSproposed by Tramel and Fowler [22] On the basis of theabove research we propose the CSVC system with superiorperformance in [14] In this paper we will evaluate H264DISCOVER and CSVC system in terms of rate-distortionperformance and rate-energy-distortion performance Thecodec process of CSVC system is described in detail below

31 Encoder Framework The encoder framework of CSVCsystem is shown in Figure 3 First the input video sequenceis divided into several GOPs and key frames and nonkeyframes are separated in the group Then the key framesand the nonkey frames are divided into 119870 nonoverlappingsubblocks of size 119861 times 119861 pixels and each block is arrangedin raster order as column vector of length 1198612 (1198612 = 119873)Finally the measurement matrix is constructed to calculatethe measurement vector of each block as follows

yK119896 = ΦK sdot xK119896 119896 = 1 119870yNK119896 = ΦNK sdot xNK119896 119896 = 1 119870 (1)

4 International Journal of Digital Multimedia Broadcasting

Regrouping

Reference frame selection

Combination

Intrareconstruction

SI prediction

Residual recovery

Unpackaging

Measurement matrix construction

Reconstructed video sequence Joint reconstruction

Nonkey frame

SI

Packet

Keyframe

GOPi

횽K

횽NK

DPCM-NQminus1 Huffmanminus1

yKk

yNKk

Figure 4 CSVC decoder architecture

where xK119896 and xNK119896 denote the 119896th subblock of the keyframe and the nonkey frame respectively andΦK119896 andΦNK119896are the measurement matrix constructed by the randomHadamard matrix For the key frames the size of themeasurementmatrixΦK119896 is119872Ktimes119873 so themeasurement rateis 119878K = 119872K119873 For nonkey frames the size of measurementmatrix ΦNK119896 is119872NK times 119873 so the measurement rate is 119878NK =119872NK119873

Then the measurement vectors yK and yNK of blocks willbe transmitted to the quantizer to form bit stream CSVCsystem uses the nonuniform quantizer based on DifferentialPulse-Code Modulation (DPCM) DPCM-NQ for shortwhich first computes the residual of measurement valuebetween adjacent subblocks to reduce coding redundancyand then quantifies the residual In consideration of thehigh frequency of small residual value the nonuniformquantization is used to process the residual of each subblockin order to reduce the quantization error Supposing that the119898thmeasurement residual of the 119896th subblock is119889119896(119898) it canbe compressed according to 120583 law as follows

119889comp = 119891 [119889119896 (119898)]= sgn [119889119896 (119898)] sdot log (1 + 120583

1003816100381610038161003816119889119896 (119898) 1198631003816100381610038161003816)log (1 + 120583) (2)

where 119863 is the maximum measurement residual of thecurrent frame sgn[sdot] represents the symbolic function and 120583is 10 In the inverse quantization the estimated value of 119889119896(119898)is calculated using the following decompression formula

119889119896 (119898) = 119891minus1 (119889comp)

= sgn [119889comp] sdot 119863120583 sdot [(1 + 120583)119889comp minus 1]

(3)

After the residual of each block is quantized the quantizeddata of all the blocks are subjected to Huffman coding andare encapsulated into data packets to be sent to the decoderside

32 Decoder Framework The decoder framework of CSVCsystem is shown in Figure 4 After the data packets are

received on the decoder side the measurement vectors yK119896of key frames and yNK119896 of nonkey frames can be obtained byHuffman decoding and inverse quantization For key framesthe intraframe reconstruction model is used as follows

xK = argminx 1003817100381710038171003817yK minusΘKE sdot x10038171003817100381710038172 + 120582 Ψ sdot x1 (4)

where

yK =[[[[[[[[[

yK1yK2

yK119870

]]]]]]]]]

ΘK =[[[[[[[[

ΦK 0ΦK

ΦK

0 ΦK

]]]]]]]]

E sdot x =[[[[[[[[[

xK1xK2

xK119870

]]]]]]]]]

(5)

Ψ is the sparse transform matrix of the video frame x and 120582represents the regularization factorThe reconstructedmodel(4) can be solved by a variety of still image CS reconstructionalgorithms To ensure high-quality recovery of key framesCSVC system uses multihypothesis smoothing Landweberiterative algorithm used in [22] to solve model (4)

For the nonkey frame we firstly obtain the side infor-mation xSI of the current nonkey frame by carrying out theside information prediction of the adjacent reconstructed keyframe and then calculate the residual measurement vector

International Journal of Digital Multimedia Broadcasting 5

between the measurement vector of each block and its sideinformation as follows

yR =[[[[[[[

yR1yR2

yR119870

]]]]]]]

=[[[[[[[

yNK1 minusΦNKxSI1yNK2 minusΦNKxSI2

yNK119870 minusΦNKxSI119870

]]]]]]]

=[[[[[[[

ΦNK sdot (xNK1 minus xSI1)ΦNK sdot (xNK2 minus xSI2)

ΦNK sdot (xNK119870 minus xSI119870)

]]]]]]]

(6)

where

ΘNK =[[[[[[

ΦNK 0ΦNK

d

0 ΦNK

]]]]]]

rNK119894 = xNK119894 minus xSI119894

(7)

So (6) can be transformed into

yR = ΘNKE sdot rNK (8)

where rNK is the residual between nonkey frame xNK and sideinformation xSI According to (8) the residual reconstructionmodel of nonkey frame can be established as follows

rNK = argminr1003817100381710038171003817yR minusΘNKE sdot r10038171003817100381710038172 + 120578 P sdot r1 (9)

where P is the sparse transform matrix of the residualrNK and 120578 denotes the regularization factor The residualreconstruction model (9) is still solved using Landweberiterative algorithm Finally the reconstruction of nonkeyframe can be calculated as follows

xNK = xSI + rNK (10)

The features of CSVC system constructed according tothe above codec process are as follows (1) compared toDISCOVER CSVC system eliminates virtual channel andfeedback channel and thus the difficulty of engineering isreduced (2) since there is no correlation between the CSmeasurement and image content the code rate is determinedonly by the measurement rate which makes it easier forCSVC system to control the code rate (3) each measurementvalue contains all the image information therefore it is easyto implement scalable coding (4) the data security can beenhanced by the random generation of the measurementmatrix The above features endow CSVC system with moreengineering value and make it become a potential newDVC scheme We are more concerned about the comparisonof coding energy consumption between CSVC system andH264 andDISCOVER so in the experiment part the codingenergy consumption of the three systems will be evaluated indetail

4 Experimental Results and Analysis

The performances of H264 DISCOVER and CSVC areevaluated respectively using four standard video sequencesnamed Foreman Bus Mobile and Football in CIF for-mat H264 adopts the standard coding configuration ofJM190 model and implements intramode DISCOVER usesthe default encoding configuration and CSVC adopts theexperimental parameter configuration in [14] The rate-distortion and rate-energy-distortion performances of thethree encoders are compared where rate-distortion reflectsthe relationship between the code rate and the Peak Signal-Noise Ratio (PSNR) while rate-energy-distortion reflects therelationship between the coding time and PSNR Using thesame experimental platform the coding time is proportionalto the energy consumption therefore it can represent thelevel of coding energy consumption The experimental plat-form is MATALB R2012b the computer system is 64-bitWindows 7 operating system with an installation memory of800GB and Intel Core i7-4900 processor whose frequency is360GHz

41 Evaluation on Rate-Distortion Performance Figure 5shows the rate-distortion curves for H264 DISCOVER andCSVC encoders under different test video sequences It canbe seen from Figure 5 that the PSNR values of the wholereconstructed video processed by different encoder alwaysgrow in a positive trend when the code rates increase Onthe whole the encoding effects of H264 and DISCOVER arealways better than CSVC encoder For the test videos Busand Football at the same code rate the video reconstructioneffect of H264 is the best for Foreman and Mobile at thesame code rate and in the specific code rate range thecoding effect of DISCOVER is even better than H264 Itcan be seen that the rate-distortion performance of H264 isoptimal DISCOVER follows and CSVC is the worst and hasa big performance difference from the other two For CSVCthe measurement rate determines the bitrate When themeasurement rate is 005 the bitrate is about 6000 kbitss Ifwe further decrease the measurement rate the bitrate will belower than 6000 kbitss The average PSNR of reconstructedvideo gradually decreases with the measurement rate linearlydecreasing The variation of PSNR curve is smooth and thePSNR value cannot suddenly reduce when the bitrate dropsto below 6000 kbitss

42 Evaluation on Rate-Energy-Distortion Performance Fig-ure 6 shows the rate-energy-distortion curves for H264DISCOVER and CSVC encoders under different test videosequences It can be seen from Figure 6 that for any videounder the same PSNR value the encoding time of CSVC isthe shortest DISCOVER follows and H264 is the longestIn particular the average encoding time of CSVC is onlyabout 3 seconds which means that the energy consumptionof CSVC is much lower than DISCOVER and H264 on thesame recovery level With the PSNR value of reconstructedvideo increasing the encoder time of H264 and DISCOVERgradually increases But the change of encoder time underDISCOVER framework is steeper and H264 framework

6 International Journal of Digital Multimedia Broadcasting

H264DISCOVERCSVC

1000 2000 3000 4000 5000 6000 7000 8000 9000 100000Bitrate (kbitss)

30

35

40

45

50

55

60PS

NR

(dB)

(a) Foreman

2000 4000 6000 8000 10000 120000Bitrate (kbitss)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(b) Bus

H264DISCOVERCSVC

2000 4000 6000 8000 10000 120000Bitrate (kbitss)

25

30

35

40

45

50

55

60

PSN

R (d

B)

(c) Football

25

30

35

40

45

50

55

PSN

R (d

B)

5000 10000 150000Bitrate (kbitss)

H264DISCOVERCSVC

(d) Mobile

Figure 5 Comparison of rate-distortion performance of H264 DISCOVER and CSVC encoder under different test video sequences

more gentle which shows that H264 has a high dependencyon energy consumption with its promotion of performanceand DISCOVER also needs a certain amount of energy inputThe computational complexity of CS measuring determinesthe energy consumption at encoder Suppose119872 denotes thenumber of CSmeasurements for a video frame and 119871 denotesthe total number of pixels in a video frame The computa-tional complexity of CS measuring is 119874(119872119871) Because119872 isfar below 119871 the variation of energy consumption is very smallwhen changing 119872 However 119872 is an important factor forthe reconstruction quality of video frameThe reconstructionquality can be improved effectively with small increments of119872 Therefore the slope of the rate-energy-distortion curveis almost vertical indicating that the small investment ofenergy consumption can get the significant improvement ofreconstruction quality It can be seen that the rate-energy-distortion performance of CSVC is optimal DISCOVERfollows H264 is the worst Among the three the energyconsumption of CSVC is approximately invariant regardlessof reconstruction quality while the reconstruction quality ofH264 andDISCOVERhas a great correlationwith the energyconsumption of coding

5 Conclusions

This paper has conducted an experiment-driven analysis ofrate-distortion and rate-energy-distortion performances ofCSVC algorithm and compares them with that of H264 andDISCOVER The rate-distortion and rate-energy-distortionperformances of the three systems are evaluated under thesame experimental environment Experiment results showthat the rate-distortion performance of CSVC has a largeperformance difference from H264 and DISCOVER but itsrate-energy-distortion performance has a greater advantagethat is the rapid improvement of its reconstruction qualitydoes not depend on coding energy input Therefore onthe premise that communication bandwidth is effectivelyimproved CSVC can be used as a candidate for futurewireless video communication because of its characteristicswhich provides wireless video terminals limited by energyconsumption and computing power with more possibili-ties

At present the rate-distortion performance of CSVC isstill not ideal and there is still some way to go before weput CSVC into practical use Efforts should be made in the

International Journal of Digital Multimedia Broadcasting 7

30

35

40

45

50

55

60PS

NR

(dB)

50 100 150 200 250 300 3500Encoder time (s)

H264DISCOVERCSVC

(a) Foreman

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(b) Bus

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

60

PSN

R (d

B)

H264DISCOVERCSVC

(c) Football

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(d) Mobile

Figure 6 Comparison of rate-energy-distortion performance ofH264 DISCOVER andCSVC encoder under different test video sequences

following areas to improve the rate-distortion performanceof CSVC

(1) Side Information Estimation Rate-distortion performanceof CSVC is greatly related to the accuracy of side informationestimation which means that high-quality side informationimmensely reduces the required supply of bit load fromencoder side Therefore finding the appropriate motion esti-mation algorithm to obtain more accurate side informationand realizing the optimal reconstruction of decoding willbecome the key to improving the rate-distortion performanceof CSVC

(2) A Priori Structural Feature Modeling of Video FramesImages of the same type often have similar structuralinformation Therefore the reconstructed model can beconstructed by evaluating the structural information of thedecoded video frames and extracting the prior knowledgewhich can reduce code rate and improve the reconstructionquality For example the statistical correlation structurecan be used in the image transformation coefficient and atree structure can be adopted for the wavelet coefficientHowever due to the complexity and uncertainty of natural

images further study should be made on how to use a prioriknowledge to construct a suitable model

(3) Quantization Measurement Uniform quantization is themajor method adopted to quantify the CS measurementcurrently But the traditional entropy coding method isnot ideal for compressing the uniform quantization valuesbecause of the statistical independence between the uniformquantization values Then how to express the CS value withthe least number of bits with the constraint of information-theoretic rate-distortion coding theorem is one of the keytopics of the following research Therefore it is necessary topropose a new nonuniform quantization method to establishstatistical correlation between quantization values and herebyto design a new entropy coding method matching thestatistical correlation

The above-mentioned further researches are employedat decoder to improve the rate-distortion performance ofCSVC but the CS measuring at encoder guarantees theadvantage of CSVC in rate-energy-distortion performanceTherefore the rate-energy-distortion performance cannot beaffected while improving the rate-distortion performance ofCSVC

8 International Journal of Digital Multimedia Broadcasting

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work was supported in part by the National NaturalScience Foundation of China under Grant no 61501393in part by the Key Scientific Research Project of Collegesand Universities in Henan Province of China under Grant16A520069 in part by Youth Sustentation Fund of XinyangNormal University under Grant no 2015-QN-043 and inpart by Scientific Research Foundation of Graduate School ofXinyang Normal University under Grant no 2016KYJJ10

References

[1] J Ostermann J Bormans P List et al ldquoVideo coding withH264AVC tools performance and complexityrdquo IEEE Circuitsand Systems Magazine vol 4 no 1 pp 7ndash28 2004

[2] B Girod A M Aaron S Rane and D Rebollo-MonederoldquoDistributed video codingrdquo Proceedings of the IEEE vol 93 no1 pp 71ndash83 2005

[3] A DWyner and J Ziv ldquoThe rate-distortion function for sourcecoding with side information at the decoderrdquo IEEE Transactionson Information Theory vol 22 no 1 pp 1ndash10 1976

[4] A Aaron S Rane R Zhang and B Girod ldquoWyner-Ziv codingfor video applications to compression and error resiliencerdquo inProceedings of the Data Compression Conference (DCC rsquo03) pp93ndash102 Snowbird Utah USA March 2003

[5] R Puri A Majumdar and K Ramchandran ldquoPRISM a videocoding paradigm with motion estimation at the decoderrdquo IEEETransactions on Image Processing vol 16 no 10 pp 2436ndash24482007

[6] Q Xu and Z Xiong ldquoLayered Wyner-Ziv video codingrdquo inProceedings of the Visual Communications and Image Processing2004 pp 83ndash91 IEEE San Jose Calif USA January 2004

[7] W Liu L Dong and W Zeng ldquoMotion refinement basedprogressive side-information estimation for Wyner-Ziv videocodingrdquo IEEE Transactions on Circuits amp Systems for VideoTechnology vol 20 no 12 pp 1863ndash1875 2010

[8] X Artigas J Ascenso M Dalai S Klomp D Kubasov and MOuaret ldquoThe discover codec architecture techniques and eval-uationrdquo in Proceedings of the 26th Picture Coding Symposium(PCS rsquo07) pp 1103ndash1120 Lisbon Portugal November 2007

[9] J Slowack J Skorupa N Deligiannis P Lambert AMunteanuand R van de Walle ldquoDistributed video coding with feedbackchannel constraintsrdquo IEEE Transactions on Circuits amp Systemsfor Video Technology vol 22 no 7 pp 1014ndash1026 2012

[10] C Brites andF Pereira ldquoCorrelationnoisemodeling for efficientpixel and transform domain Wyner-Ziv video codingrdquo IEEETransactions on Circuits and Systems for Video Technology vol18 no 9 pp 1177ndash1190 2008

[11] E J Candes J Romberg and T Tao ldquoRobust uncertaintyprinciples exact signal reconstruction from highly incompletefrequency informationrdquo IEEE Transactions on InformationThe-ory vol 52 no 2 pp 489ndash509 2006

[12] D L Donoho ldquoCompressed sensingrdquo IEEE Transactions onInformation Theory vol 52 no 4 pp 1289ndash1306 2006

[13] E J Candes and M B Wakin ldquoAn introduction to compressivesampling a sensingsampling paradigm that goes against thecommon knowledge in data acquisitionrdquo IEEE Signal ProcessingMagazine vol 25 no 2 pp 21ndash30 2008

[14] R Li H Liu R Xue and Y Li ldquoCompressive-sensing-basedvideo codec by autoregressive prediction and adaptive residualrecoveryrdquo International Journal of Distributed Sensor Networksvol 2015 Article ID 562840 19 pages 2015

[15] Y Liu X Zhu L Zhang and S H Cho ldquoDistributed com-pressed video sensing in camera sensor networksrdquo InternationalJournal of Distributed Sensor Networks vol 2012 Article ID352167 10 pages 2012

[16] C Di Laura D Pajuelo andG Kemper ldquoA novel steganographytechnique for SDTV-H264AVC encoded videordquo InternationalJournal of Digital Multimedia Broadcasting vol 2016 Article ID6950592 9 pages 2016

[17] Y Lahbabi E H Ibn Elhaj and A Hammouch ldquoAdaptivestreaming of scalable videos over P2PTVrdquo International Journalof Digital Multimedia Broadcasting vol 2015 Article ID 28309710 pages 2015

[18] S Pudlewski and T Melodia ldquoCompressive video streamingdesign and rate-energy-distortion analysisrdquo IEEE Transactionson Multimedia vol 15 no 8 pp 2072ndash2086 2013

[19] D Baron M B Wakin M F Duarte S Sarvotham and R GBaraniuk ldquoDistributed compressive sensingrdquo httparxivorgabs09013403

[20] J Prades-Nebot Y Ma and T Huang ldquoDistributed videocoding using compressive samplingrdquo in Proceedings of thePicture Coding Symposium (PCS rsquo09) pp 1ndash4 Chicago Ill USAMay 2009

[21] T T Do Y Chen D T Nguyen N Nguyen L Gan and T DTran ldquoDistributed compressed video sensingrdquo in Proceedingsof the 16th IEEE International Conference on Image Processing(ICIP rsquo09) pp 1393ndash1396 IEEE Cairo Egypt November 2009

[22] E W Tramel and J E Fowler ldquoVideo compressed sensingwith multihypothesisrdquo in Proceedings of the Data CompressionConference (DCC rsquo11) pp 193ndash202 Snowbird Utah USAMarch 2011

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpswwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 4: Rate-Distortion and Rate-Energy-Distortion Evaluations of ...

4 International Journal of Digital Multimedia Broadcasting

Regrouping

Reference frame selection

Combination

Intrareconstruction

SI prediction

Residual recovery

Unpackaging

Measurement matrix construction

Reconstructed video sequence Joint reconstruction

Nonkey frame

SI

Packet

Keyframe

GOPi

횽K

횽NK

DPCM-NQminus1 Huffmanminus1

yKk

yNKk

Figure 4 CSVC decoder architecture

where xK119896 and xNK119896 denote the 119896th subblock of the keyframe and the nonkey frame respectively andΦK119896 andΦNK119896are the measurement matrix constructed by the randomHadamard matrix For the key frames the size of themeasurementmatrixΦK119896 is119872Ktimes119873 so themeasurement rateis 119878K = 119872K119873 For nonkey frames the size of measurementmatrix ΦNK119896 is119872NK times 119873 so the measurement rate is 119878NK =119872NK119873

Then the measurement vectors yK and yNK of blocks willbe transmitted to the quantizer to form bit stream CSVCsystem uses the nonuniform quantizer based on DifferentialPulse-Code Modulation (DPCM) DPCM-NQ for shortwhich first computes the residual of measurement valuebetween adjacent subblocks to reduce coding redundancyand then quantifies the residual In consideration of thehigh frequency of small residual value the nonuniformquantization is used to process the residual of each subblockin order to reduce the quantization error Supposing that the119898thmeasurement residual of the 119896th subblock is119889119896(119898) it canbe compressed according to 120583 law as follows

119889comp = 119891 [119889119896 (119898)]= sgn [119889119896 (119898)] sdot log (1 + 120583

1003816100381610038161003816119889119896 (119898) 1198631003816100381610038161003816)log (1 + 120583) (2)

where 119863 is the maximum measurement residual of thecurrent frame sgn[sdot] represents the symbolic function and 120583is 10 In the inverse quantization the estimated value of 119889119896(119898)is calculated using the following decompression formula

119889119896 (119898) = 119891minus1 (119889comp)

= sgn [119889comp] sdot 119863120583 sdot [(1 + 120583)119889comp minus 1]

(3)

After the residual of each block is quantized the quantizeddata of all the blocks are subjected to Huffman coding andare encapsulated into data packets to be sent to the decoderside

32 Decoder Framework The decoder framework of CSVCsystem is shown in Figure 4 After the data packets are

received on the decoder side the measurement vectors yK119896of key frames and yNK119896 of nonkey frames can be obtained byHuffman decoding and inverse quantization For key framesthe intraframe reconstruction model is used as follows

xK = argminx 1003817100381710038171003817yK minusΘKE sdot x10038171003817100381710038172 + 120582 Ψ sdot x1 (4)

where

yK =[[[[[[[[[

yK1yK2

yK119870

]]]]]]]]]

ΘK =[[[[[[[[

ΦK 0ΦK

ΦK

0 ΦK

]]]]]]]]

E sdot x =[[[[[[[[[

xK1xK2

xK119870

]]]]]]]]]

(5)

Ψ is the sparse transform matrix of the video frame x and 120582represents the regularization factorThe reconstructedmodel(4) can be solved by a variety of still image CS reconstructionalgorithms To ensure high-quality recovery of key framesCSVC system uses multihypothesis smoothing Landweberiterative algorithm used in [22] to solve model (4)

For the nonkey frame we firstly obtain the side infor-mation xSI of the current nonkey frame by carrying out theside information prediction of the adjacent reconstructed keyframe and then calculate the residual measurement vector

International Journal of Digital Multimedia Broadcasting 5

between the measurement vector of each block and its sideinformation as follows

yR =[[[[[[[

yR1yR2

yR119870

]]]]]]]

=[[[[[[[

yNK1 minusΦNKxSI1yNK2 minusΦNKxSI2

yNK119870 minusΦNKxSI119870

]]]]]]]

=[[[[[[[

ΦNK sdot (xNK1 minus xSI1)ΦNK sdot (xNK2 minus xSI2)

ΦNK sdot (xNK119870 minus xSI119870)

]]]]]]]

(6)

where

ΘNK =[[[[[[

ΦNK 0ΦNK

d

0 ΦNK

]]]]]]

rNK119894 = xNK119894 minus xSI119894

(7)

So (6) can be transformed into

yR = ΘNKE sdot rNK (8)

where rNK is the residual between nonkey frame xNK and sideinformation xSI According to (8) the residual reconstructionmodel of nonkey frame can be established as follows

rNK = argminr1003817100381710038171003817yR minusΘNKE sdot r10038171003817100381710038172 + 120578 P sdot r1 (9)

where P is the sparse transform matrix of the residualrNK and 120578 denotes the regularization factor The residualreconstruction model (9) is still solved using Landweberiterative algorithm Finally the reconstruction of nonkeyframe can be calculated as follows

xNK = xSI + rNK (10)

The features of CSVC system constructed according tothe above codec process are as follows (1) compared toDISCOVER CSVC system eliminates virtual channel andfeedback channel and thus the difficulty of engineering isreduced (2) since there is no correlation between the CSmeasurement and image content the code rate is determinedonly by the measurement rate which makes it easier forCSVC system to control the code rate (3) each measurementvalue contains all the image information therefore it is easyto implement scalable coding (4) the data security can beenhanced by the random generation of the measurementmatrix The above features endow CSVC system with moreengineering value and make it become a potential newDVC scheme We are more concerned about the comparisonof coding energy consumption between CSVC system andH264 andDISCOVER so in the experiment part the codingenergy consumption of the three systems will be evaluated indetail

4 Experimental Results and Analysis

The performances of H264 DISCOVER and CSVC areevaluated respectively using four standard video sequencesnamed Foreman Bus Mobile and Football in CIF for-mat H264 adopts the standard coding configuration ofJM190 model and implements intramode DISCOVER usesthe default encoding configuration and CSVC adopts theexperimental parameter configuration in [14] The rate-distortion and rate-energy-distortion performances of thethree encoders are compared where rate-distortion reflectsthe relationship between the code rate and the Peak Signal-Noise Ratio (PSNR) while rate-energy-distortion reflects therelationship between the coding time and PSNR Using thesame experimental platform the coding time is proportionalto the energy consumption therefore it can represent thelevel of coding energy consumption The experimental plat-form is MATALB R2012b the computer system is 64-bitWindows 7 operating system with an installation memory of800GB and Intel Core i7-4900 processor whose frequency is360GHz

41 Evaluation on Rate-Distortion Performance Figure 5shows the rate-distortion curves for H264 DISCOVER andCSVC encoders under different test video sequences It canbe seen from Figure 5 that the PSNR values of the wholereconstructed video processed by different encoder alwaysgrow in a positive trend when the code rates increase Onthe whole the encoding effects of H264 and DISCOVER arealways better than CSVC encoder For the test videos Busand Football at the same code rate the video reconstructioneffect of H264 is the best for Foreman and Mobile at thesame code rate and in the specific code rate range thecoding effect of DISCOVER is even better than H264 Itcan be seen that the rate-distortion performance of H264 isoptimal DISCOVER follows and CSVC is the worst and hasa big performance difference from the other two For CSVCthe measurement rate determines the bitrate When themeasurement rate is 005 the bitrate is about 6000 kbitss Ifwe further decrease the measurement rate the bitrate will belower than 6000 kbitss The average PSNR of reconstructedvideo gradually decreases with the measurement rate linearlydecreasing The variation of PSNR curve is smooth and thePSNR value cannot suddenly reduce when the bitrate dropsto below 6000 kbitss

42 Evaluation on Rate-Energy-Distortion Performance Fig-ure 6 shows the rate-energy-distortion curves for H264DISCOVER and CSVC encoders under different test videosequences It can be seen from Figure 6 that for any videounder the same PSNR value the encoding time of CSVC isthe shortest DISCOVER follows and H264 is the longestIn particular the average encoding time of CSVC is onlyabout 3 seconds which means that the energy consumptionof CSVC is much lower than DISCOVER and H264 on thesame recovery level With the PSNR value of reconstructedvideo increasing the encoder time of H264 and DISCOVERgradually increases But the change of encoder time underDISCOVER framework is steeper and H264 framework

6 International Journal of Digital Multimedia Broadcasting

H264DISCOVERCSVC

1000 2000 3000 4000 5000 6000 7000 8000 9000 100000Bitrate (kbitss)

30

35

40

45

50

55

60PS

NR

(dB)

(a) Foreman

2000 4000 6000 8000 10000 120000Bitrate (kbitss)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(b) Bus

H264DISCOVERCSVC

2000 4000 6000 8000 10000 120000Bitrate (kbitss)

25

30

35

40

45

50

55

60

PSN

R (d

B)

(c) Football

25

30

35

40

45

50

55

PSN

R (d

B)

5000 10000 150000Bitrate (kbitss)

H264DISCOVERCSVC

(d) Mobile

Figure 5 Comparison of rate-distortion performance of H264 DISCOVER and CSVC encoder under different test video sequences

more gentle which shows that H264 has a high dependencyon energy consumption with its promotion of performanceand DISCOVER also needs a certain amount of energy inputThe computational complexity of CS measuring determinesthe energy consumption at encoder Suppose119872 denotes thenumber of CSmeasurements for a video frame and 119871 denotesthe total number of pixels in a video frame The computa-tional complexity of CS measuring is 119874(119872119871) Because119872 isfar below 119871 the variation of energy consumption is very smallwhen changing 119872 However 119872 is an important factor forthe reconstruction quality of video frameThe reconstructionquality can be improved effectively with small increments of119872 Therefore the slope of the rate-energy-distortion curveis almost vertical indicating that the small investment ofenergy consumption can get the significant improvement ofreconstruction quality It can be seen that the rate-energy-distortion performance of CSVC is optimal DISCOVERfollows H264 is the worst Among the three the energyconsumption of CSVC is approximately invariant regardlessof reconstruction quality while the reconstruction quality ofH264 andDISCOVERhas a great correlationwith the energyconsumption of coding

5 Conclusions

This paper has conducted an experiment-driven analysis ofrate-distortion and rate-energy-distortion performances ofCSVC algorithm and compares them with that of H264 andDISCOVER The rate-distortion and rate-energy-distortionperformances of the three systems are evaluated under thesame experimental environment Experiment results showthat the rate-distortion performance of CSVC has a largeperformance difference from H264 and DISCOVER but itsrate-energy-distortion performance has a greater advantagethat is the rapid improvement of its reconstruction qualitydoes not depend on coding energy input Therefore onthe premise that communication bandwidth is effectivelyimproved CSVC can be used as a candidate for futurewireless video communication because of its characteristicswhich provides wireless video terminals limited by energyconsumption and computing power with more possibili-ties

At present the rate-distortion performance of CSVC isstill not ideal and there is still some way to go before weput CSVC into practical use Efforts should be made in the

International Journal of Digital Multimedia Broadcasting 7

30

35

40

45

50

55

60PS

NR

(dB)

50 100 150 200 250 300 3500Encoder time (s)

H264DISCOVERCSVC

(a) Foreman

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(b) Bus

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

60

PSN

R (d

B)

H264DISCOVERCSVC

(c) Football

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(d) Mobile

Figure 6 Comparison of rate-energy-distortion performance ofH264 DISCOVER andCSVC encoder under different test video sequences

following areas to improve the rate-distortion performanceof CSVC

(1) Side Information Estimation Rate-distortion performanceof CSVC is greatly related to the accuracy of side informationestimation which means that high-quality side informationimmensely reduces the required supply of bit load fromencoder side Therefore finding the appropriate motion esti-mation algorithm to obtain more accurate side informationand realizing the optimal reconstruction of decoding willbecome the key to improving the rate-distortion performanceof CSVC

(2) A Priori Structural Feature Modeling of Video FramesImages of the same type often have similar structuralinformation Therefore the reconstructed model can beconstructed by evaluating the structural information of thedecoded video frames and extracting the prior knowledgewhich can reduce code rate and improve the reconstructionquality For example the statistical correlation structurecan be used in the image transformation coefficient and atree structure can be adopted for the wavelet coefficientHowever due to the complexity and uncertainty of natural

images further study should be made on how to use a prioriknowledge to construct a suitable model

(3) Quantization Measurement Uniform quantization is themajor method adopted to quantify the CS measurementcurrently But the traditional entropy coding method isnot ideal for compressing the uniform quantization valuesbecause of the statistical independence between the uniformquantization values Then how to express the CS value withthe least number of bits with the constraint of information-theoretic rate-distortion coding theorem is one of the keytopics of the following research Therefore it is necessary topropose a new nonuniform quantization method to establishstatistical correlation between quantization values and herebyto design a new entropy coding method matching thestatistical correlation

The above-mentioned further researches are employedat decoder to improve the rate-distortion performance ofCSVC but the CS measuring at encoder guarantees theadvantage of CSVC in rate-energy-distortion performanceTherefore the rate-energy-distortion performance cannot beaffected while improving the rate-distortion performance ofCSVC

8 International Journal of Digital Multimedia Broadcasting

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work was supported in part by the National NaturalScience Foundation of China under Grant no 61501393in part by the Key Scientific Research Project of Collegesand Universities in Henan Province of China under Grant16A520069 in part by Youth Sustentation Fund of XinyangNormal University under Grant no 2015-QN-043 and inpart by Scientific Research Foundation of Graduate School ofXinyang Normal University under Grant no 2016KYJJ10

References

[1] J Ostermann J Bormans P List et al ldquoVideo coding withH264AVC tools performance and complexityrdquo IEEE Circuitsand Systems Magazine vol 4 no 1 pp 7ndash28 2004

[2] B Girod A M Aaron S Rane and D Rebollo-MonederoldquoDistributed video codingrdquo Proceedings of the IEEE vol 93 no1 pp 71ndash83 2005

[3] A DWyner and J Ziv ldquoThe rate-distortion function for sourcecoding with side information at the decoderrdquo IEEE Transactionson Information Theory vol 22 no 1 pp 1ndash10 1976

[4] A Aaron S Rane R Zhang and B Girod ldquoWyner-Ziv codingfor video applications to compression and error resiliencerdquo inProceedings of the Data Compression Conference (DCC rsquo03) pp93ndash102 Snowbird Utah USA March 2003

[5] R Puri A Majumdar and K Ramchandran ldquoPRISM a videocoding paradigm with motion estimation at the decoderrdquo IEEETransactions on Image Processing vol 16 no 10 pp 2436ndash24482007

[6] Q Xu and Z Xiong ldquoLayered Wyner-Ziv video codingrdquo inProceedings of the Visual Communications and Image Processing2004 pp 83ndash91 IEEE San Jose Calif USA January 2004

[7] W Liu L Dong and W Zeng ldquoMotion refinement basedprogressive side-information estimation for Wyner-Ziv videocodingrdquo IEEE Transactions on Circuits amp Systems for VideoTechnology vol 20 no 12 pp 1863ndash1875 2010

[8] X Artigas J Ascenso M Dalai S Klomp D Kubasov and MOuaret ldquoThe discover codec architecture techniques and eval-uationrdquo in Proceedings of the 26th Picture Coding Symposium(PCS rsquo07) pp 1103ndash1120 Lisbon Portugal November 2007

[9] J Slowack J Skorupa N Deligiannis P Lambert AMunteanuand R van de Walle ldquoDistributed video coding with feedbackchannel constraintsrdquo IEEE Transactions on Circuits amp Systemsfor Video Technology vol 22 no 7 pp 1014ndash1026 2012

[10] C Brites andF Pereira ldquoCorrelationnoisemodeling for efficientpixel and transform domain Wyner-Ziv video codingrdquo IEEETransactions on Circuits and Systems for Video Technology vol18 no 9 pp 1177ndash1190 2008

[11] E J Candes J Romberg and T Tao ldquoRobust uncertaintyprinciples exact signal reconstruction from highly incompletefrequency informationrdquo IEEE Transactions on InformationThe-ory vol 52 no 2 pp 489ndash509 2006

[12] D L Donoho ldquoCompressed sensingrdquo IEEE Transactions onInformation Theory vol 52 no 4 pp 1289ndash1306 2006

[13] E J Candes and M B Wakin ldquoAn introduction to compressivesampling a sensingsampling paradigm that goes against thecommon knowledge in data acquisitionrdquo IEEE Signal ProcessingMagazine vol 25 no 2 pp 21ndash30 2008

[14] R Li H Liu R Xue and Y Li ldquoCompressive-sensing-basedvideo codec by autoregressive prediction and adaptive residualrecoveryrdquo International Journal of Distributed Sensor Networksvol 2015 Article ID 562840 19 pages 2015

[15] Y Liu X Zhu L Zhang and S H Cho ldquoDistributed com-pressed video sensing in camera sensor networksrdquo InternationalJournal of Distributed Sensor Networks vol 2012 Article ID352167 10 pages 2012

[16] C Di Laura D Pajuelo andG Kemper ldquoA novel steganographytechnique for SDTV-H264AVC encoded videordquo InternationalJournal of Digital Multimedia Broadcasting vol 2016 Article ID6950592 9 pages 2016

[17] Y Lahbabi E H Ibn Elhaj and A Hammouch ldquoAdaptivestreaming of scalable videos over P2PTVrdquo International Journalof Digital Multimedia Broadcasting vol 2015 Article ID 28309710 pages 2015

[18] S Pudlewski and T Melodia ldquoCompressive video streamingdesign and rate-energy-distortion analysisrdquo IEEE Transactionson Multimedia vol 15 no 8 pp 2072ndash2086 2013

[19] D Baron M B Wakin M F Duarte S Sarvotham and R GBaraniuk ldquoDistributed compressive sensingrdquo httparxivorgabs09013403

[20] J Prades-Nebot Y Ma and T Huang ldquoDistributed videocoding using compressive samplingrdquo in Proceedings of thePicture Coding Symposium (PCS rsquo09) pp 1ndash4 Chicago Ill USAMay 2009

[21] T T Do Y Chen D T Nguyen N Nguyen L Gan and T DTran ldquoDistributed compressed video sensingrdquo in Proceedingsof the 16th IEEE International Conference on Image Processing(ICIP rsquo09) pp 1393ndash1396 IEEE Cairo Egypt November 2009

[22] E W Tramel and J E Fowler ldquoVideo compressed sensingwith multihypothesisrdquo in Proceedings of the Data CompressionConference (DCC rsquo11) pp 193ndash202 Snowbird Utah USAMarch 2011

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpswwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 5: Rate-Distortion and Rate-Energy-Distortion Evaluations of ...

International Journal of Digital Multimedia Broadcasting 5

between the measurement vector of each block and its sideinformation as follows

yR =[[[[[[[

yR1yR2

yR119870

]]]]]]]

=[[[[[[[

yNK1 minusΦNKxSI1yNK2 minusΦNKxSI2

yNK119870 minusΦNKxSI119870

]]]]]]]

=[[[[[[[

ΦNK sdot (xNK1 minus xSI1)ΦNK sdot (xNK2 minus xSI2)

ΦNK sdot (xNK119870 minus xSI119870)

]]]]]]]

(6)

where

ΘNK =[[[[[[

ΦNK 0ΦNK

d

0 ΦNK

]]]]]]

rNK119894 = xNK119894 minus xSI119894

(7)

So (6) can be transformed into

yR = ΘNKE sdot rNK (8)

where rNK is the residual between nonkey frame xNK and sideinformation xSI According to (8) the residual reconstructionmodel of nonkey frame can be established as follows

rNK = argminr1003817100381710038171003817yR minusΘNKE sdot r10038171003817100381710038172 + 120578 P sdot r1 (9)

where P is the sparse transform matrix of the residualrNK and 120578 denotes the regularization factor The residualreconstruction model (9) is still solved using Landweberiterative algorithm Finally the reconstruction of nonkeyframe can be calculated as follows

xNK = xSI + rNK (10)

The features of CSVC system constructed according tothe above codec process are as follows (1) compared toDISCOVER CSVC system eliminates virtual channel andfeedback channel and thus the difficulty of engineering isreduced (2) since there is no correlation between the CSmeasurement and image content the code rate is determinedonly by the measurement rate which makes it easier forCSVC system to control the code rate (3) each measurementvalue contains all the image information therefore it is easyto implement scalable coding (4) the data security can beenhanced by the random generation of the measurementmatrix The above features endow CSVC system with moreengineering value and make it become a potential newDVC scheme We are more concerned about the comparisonof coding energy consumption between CSVC system andH264 andDISCOVER so in the experiment part the codingenergy consumption of the three systems will be evaluated indetail

4 Experimental Results and Analysis

The performances of H264 DISCOVER and CSVC areevaluated respectively using four standard video sequencesnamed Foreman Bus Mobile and Football in CIF for-mat H264 adopts the standard coding configuration ofJM190 model and implements intramode DISCOVER usesthe default encoding configuration and CSVC adopts theexperimental parameter configuration in [14] The rate-distortion and rate-energy-distortion performances of thethree encoders are compared where rate-distortion reflectsthe relationship between the code rate and the Peak Signal-Noise Ratio (PSNR) while rate-energy-distortion reflects therelationship between the coding time and PSNR Using thesame experimental platform the coding time is proportionalto the energy consumption therefore it can represent thelevel of coding energy consumption The experimental plat-form is MATALB R2012b the computer system is 64-bitWindows 7 operating system with an installation memory of800GB and Intel Core i7-4900 processor whose frequency is360GHz

41 Evaluation on Rate-Distortion Performance Figure 5shows the rate-distortion curves for H264 DISCOVER andCSVC encoders under different test video sequences It canbe seen from Figure 5 that the PSNR values of the wholereconstructed video processed by different encoder alwaysgrow in a positive trend when the code rates increase Onthe whole the encoding effects of H264 and DISCOVER arealways better than CSVC encoder For the test videos Busand Football at the same code rate the video reconstructioneffect of H264 is the best for Foreman and Mobile at thesame code rate and in the specific code rate range thecoding effect of DISCOVER is even better than H264 Itcan be seen that the rate-distortion performance of H264 isoptimal DISCOVER follows and CSVC is the worst and hasa big performance difference from the other two For CSVCthe measurement rate determines the bitrate When themeasurement rate is 005 the bitrate is about 6000 kbitss Ifwe further decrease the measurement rate the bitrate will belower than 6000 kbitss The average PSNR of reconstructedvideo gradually decreases with the measurement rate linearlydecreasing The variation of PSNR curve is smooth and thePSNR value cannot suddenly reduce when the bitrate dropsto below 6000 kbitss

42 Evaluation on Rate-Energy-Distortion Performance Fig-ure 6 shows the rate-energy-distortion curves for H264DISCOVER and CSVC encoders under different test videosequences It can be seen from Figure 6 that for any videounder the same PSNR value the encoding time of CSVC isthe shortest DISCOVER follows and H264 is the longestIn particular the average encoding time of CSVC is onlyabout 3 seconds which means that the energy consumptionof CSVC is much lower than DISCOVER and H264 on thesame recovery level With the PSNR value of reconstructedvideo increasing the encoder time of H264 and DISCOVERgradually increases But the change of encoder time underDISCOVER framework is steeper and H264 framework

6 International Journal of Digital Multimedia Broadcasting

H264DISCOVERCSVC

1000 2000 3000 4000 5000 6000 7000 8000 9000 100000Bitrate (kbitss)

30

35

40

45

50

55

60PS

NR

(dB)

(a) Foreman

2000 4000 6000 8000 10000 120000Bitrate (kbitss)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(b) Bus

H264DISCOVERCSVC

2000 4000 6000 8000 10000 120000Bitrate (kbitss)

25

30

35

40

45

50

55

60

PSN

R (d

B)

(c) Football

25

30

35

40

45

50

55

PSN

R (d

B)

5000 10000 150000Bitrate (kbitss)

H264DISCOVERCSVC

(d) Mobile

Figure 5 Comparison of rate-distortion performance of H264 DISCOVER and CSVC encoder under different test video sequences

more gentle which shows that H264 has a high dependencyon energy consumption with its promotion of performanceand DISCOVER also needs a certain amount of energy inputThe computational complexity of CS measuring determinesthe energy consumption at encoder Suppose119872 denotes thenumber of CSmeasurements for a video frame and 119871 denotesthe total number of pixels in a video frame The computa-tional complexity of CS measuring is 119874(119872119871) Because119872 isfar below 119871 the variation of energy consumption is very smallwhen changing 119872 However 119872 is an important factor forthe reconstruction quality of video frameThe reconstructionquality can be improved effectively with small increments of119872 Therefore the slope of the rate-energy-distortion curveis almost vertical indicating that the small investment ofenergy consumption can get the significant improvement ofreconstruction quality It can be seen that the rate-energy-distortion performance of CSVC is optimal DISCOVERfollows H264 is the worst Among the three the energyconsumption of CSVC is approximately invariant regardlessof reconstruction quality while the reconstruction quality ofH264 andDISCOVERhas a great correlationwith the energyconsumption of coding

5 Conclusions

This paper has conducted an experiment-driven analysis ofrate-distortion and rate-energy-distortion performances ofCSVC algorithm and compares them with that of H264 andDISCOVER The rate-distortion and rate-energy-distortionperformances of the three systems are evaluated under thesame experimental environment Experiment results showthat the rate-distortion performance of CSVC has a largeperformance difference from H264 and DISCOVER but itsrate-energy-distortion performance has a greater advantagethat is the rapid improvement of its reconstruction qualitydoes not depend on coding energy input Therefore onthe premise that communication bandwidth is effectivelyimproved CSVC can be used as a candidate for futurewireless video communication because of its characteristicswhich provides wireless video terminals limited by energyconsumption and computing power with more possibili-ties

At present the rate-distortion performance of CSVC isstill not ideal and there is still some way to go before weput CSVC into practical use Efforts should be made in the

International Journal of Digital Multimedia Broadcasting 7

30

35

40

45

50

55

60PS

NR

(dB)

50 100 150 200 250 300 3500Encoder time (s)

H264DISCOVERCSVC

(a) Foreman

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(b) Bus

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

60

PSN

R (d

B)

H264DISCOVERCSVC

(c) Football

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(d) Mobile

Figure 6 Comparison of rate-energy-distortion performance ofH264 DISCOVER andCSVC encoder under different test video sequences

following areas to improve the rate-distortion performanceof CSVC

(1) Side Information Estimation Rate-distortion performanceof CSVC is greatly related to the accuracy of side informationestimation which means that high-quality side informationimmensely reduces the required supply of bit load fromencoder side Therefore finding the appropriate motion esti-mation algorithm to obtain more accurate side informationand realizing the optimal reconstruction of decoding willbecome the key to improving the rate-distortion performanceof CSVC

(2) A Priori Structural Feature Modeling of Video FramesImages of the same type often have similar structuralinformation Therefore the reconstructed model can beconstructed by evaluating the structural information of thedecoded video frames and extracting the prior knowledgewhich can reduce code rate and improve the reconstructionquality For example the statistical correlation structurecan be used in the image transformation coefficient and atree structure can be adopted for the wavelet coefficientHowever due to the complexity and uncertainty of natural

images further study should be made on how to use a prioriknowledge to construct a suitable model

(3) Quantization Measurement Uniform quantization is themajor method adopted to quantify the CS measurementcurrently But the traditional entropy coding method isnot ideal for compressing the uniform quantization valuesbecause of the statistical independence between the uniformquantization values Then how to express the CS value withthe least number of bits with the constraint of information-theoretic rate-distortion coding theorem is one of the keytopics of the following research Therefore it is necessary topropose a new nonuniform quantization method to establishstatistical correlation between quantization values and herebyto design a new entropy coding method matching thestatistical correlation

The above-mentioned further researches are employedat decoder to improve the rate-distortion performance ofCSVC but the CS measuring at encoder guarantees theadvantage of CSVC in rate-energy-distortion performanceTherefore the rate-energy-distortion performance cannot beaffected while improving the rate-distortion performance ofCSVC

8 International Journal of Digital Multimedia Broadcasting

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work was supported in part by the National NaturalScience Foundation of China under Grant no 61501393in part by the Key Scientific Research Project of Collegesand Universities in Henan Province of China under Grant16A520069 in part by Youth Sustentation Fund of XinyangNormal University under Grant no 2015-QN-043 and inpart by Scientific Research Foundation of Graduate School ofXinyang Normal University under Grant no 2016KYJJ10

References

[1] J Ostermann J Bormans P List et al ldquoVideo coding withH264AVC tools performance and complexityrdquo IEEE Circuitsand Systems Magazine vol 4 no 1 pp 7ndash28 2004

[2] B Girod A M Aaron S Rane and D Rebollo-MonederoldquoDistributed video codingrdquo Proceedings of the IEEE vol 93 no1 pp 71ndash83 2005

[3] A DWyner and J Ziv ldquoThe rate-distortion function for sourcecoding with side information at the decoderrdquo IEEE Transactionson Information Theory vol 22 no 1 pp 1ndash10 1976

[4] A Aaron S Rane R Zhang and B Girod ldquoWyner-Ziv codingfor video applications to compression and error resiliencerdquo inProceedings of the Data Compression Conference (DCC rsquo03) pp93ndash102 Snowbird Utah USA March 2003

[5] R Puri A Majumdar and K Ramchandran ldquoPRISM a videocoding paradigm with motion estimation at the decoderrdquo IEEETransactions on Image Processing vol 16 no 10 pp 2436ndash24482007

[6] Q Xu and Z Xiong ldquoLayered Wyner-Ziv video codingrdquo inProceedings of the Visual Communications and Image Processing2004 pp 83ndash91 IEEE San Jose Calif USA January 2004

[7] W Liu L Dong and W Zeng ldquoMotion refinement basedprogressive side-information estimation for Wyner-Ziv videocodingrdquo IEEE Transactions on Circuits amp Systems for VideoTechnology vol 20 no 12 pp 1863ndash1875 2010

[8] X Artigas J Ascenso M Dalai S Klomp D Kubasov and MOuaret ldquoThe discover codec architecture techniques and eval-uationrdquo in Proceedings of the 26th Picture Coding Symposium(PCS rsquo07) pp 1103ndash1120 Lisbon Portugal November 2007

[9] J Slowack J Skorupa N Deligiannis P Lambert AMunteanuand R van de Walle ldquoDistributed video coding with feedbackchannel constraintsrdquo IEEE Transactions on Circuits amp Systemsfor Video Technology vol 22 no 7 pp 1014ndash1026 2012

[10] C Brites andF Pereira ldquoCorrelationnoisemodeling for efficientpixel and transform domain Wyner-Ziv video codingrdquo IEEETransactions on Circuits and Systems for Video Technology vol18 no 9 pp 1177ndash1190 2008

[11] E J Candes J Romberg and T Tao ldquoRobust uncertaintyprinciples exact signal reconstruction from highly incompletefrequency informationrdquo IEEE Transactions on InformationThe-ory vol 52 no 2 pp 489ndash509 2006

[12] D L Donoho ldquoCompressed sensingrdquo IEEE Transactions onInformation Theory vol 52 no 4 pp 1289ndash1306 2006

[13] E J Candes and M B Wakin ldquoAn introduction to compressivesampling a sensingsampling paradigm that goes against thecommon knowledge in data acquisitionrdquo IEEE Signal ProcessingMagazine vol 25 no 2 pp 21ndash30 2008

[14] R Li H Liu R Xue and Y Li ldquoCompressive-sensing-basedvideo codec by autoregressive prediction and adaptive residualrecoveryrdquo International Journal of Distributed Sensor Networksvol 2015 Article ID 562840 19 pages 2015

[15] Y Liu X Zhu L Zhang and S H Cho ldquoDistributed com-pressed video sensing in camera sensor networksrdquo InternationalJournal of Distributed Sensor Networks vol 2012 Article ID352167 10 pages 2012

[16] C Di Laura D Pajuelo andG Kemper ldquoA novel steganographytechnique for SDTV-H264AVC encoded videordquo InternationalJournal of Digital Multimedia Broadcasting vol 2016 Article ID6950592 9 pages 2016

[17] Y Lahbabi E H Ibn Elhaj and A Hammouch ldquoAdaptivestreaming of scalable videos over P2PTVrdquo International Journalof Digital Multimedia Broadcasting vol 2015 Article ID 28309710 pages 2015

[18] S Pudlewski and T Melodia ldquoCompressive video streamingdesign and rate-energy-distortion analysisrdquo IEEE Transactionson Multimedia vol 15 no 8 pp 2072ndash2086 2013

[19] D Baron M B Wakin M F Duarte S Sarvotham and R GBaraniuk ldquoDistributed compressive sensingrdquo httparxivorgabs09013403

[20] J Prades-Nebot Y Ma and T Huang ldquoDistributed videocoding using compressive samplingrdquo in Proceedings of thePicture Coding Symposium (PCS rsquo09) pp 1ndash4 Chicago Ill USAMay 2009

[21] T T Do Y Chen D T Nguyen N Nguyen L Gan and T DTran ldquoDistributed compressed video sensingrdquo in Proceedingsof the 16th IEEE International Conference on Image Processing(ICIP rsquo09) pp 1393ndash1396 IEEE Cairo Egypt November 2009

[22] E W Tramel and J E Fowler ldquoVideo compressed sensingwith multihypothesisrdquo in Proceedings of the Data CompressionConference (DCC rsquo11) pp 193ndash202 Snowbird Utah USAMarch 2011

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpswwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 6: Rate-Distortion and Rate-Energy-Distortion Evaluations of ...

6 International Journal of Digital Multimedia Broadcasting

H264DISCOVERCSVC

1000 2000 3000 4000 5000 6000 7000 8000 9000 100000Bitrate (kbitss)

30

35

40

45

50

55

60PS

NR

(dB)

(a) Foreman

2000 4000 6000 8000 10000 120000Bitrate (kbitss)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(b) Bus

H264DISCOVERCSVC

2000 4000 6000 8000 10000 120000Bitrate (kbitss)

25

30

35

40

45

50

55

60

PSN

R (d

B)

(c) Football

25

30

35

40

45

50

55

PSN

R (d

B)

5000 10000 150000Bitrate (kbitss)

H264DISCOVERCSVC

(d) Mobile

Figure 5 Comparison of rate-distortion performance of H264 DISCOVER and CSVC encoder under different test video sequences

more gentle which shows that H264 has a high dependencyon energy consumption with its promotion of performanceand DISCOVER also needs a certain amount of energy inputThe computational complexity of CS measuring determinesthe energy consumption at encoder Suppose119872 denotes thenumber of CSmeasurements for a video frame and 119871 denotesthe total number of pixels in a video frame The computa-tional complexity of CS measuring is 119874(119872119871) Because119872 isfar below 119871 the variation of energy consumption is very smallwhen changing 119872 However 119872 is an important factor forthe reconstruction quality of video frameThe reconstructionquality can be improved effectively with small increments of119872 Therefore the slope of the rate-energy-distortion curveis almost vertical indicating that the small investment ofenergy consumption can get the significant improvement ofreconstruction quality It can be seen that the rate-energy-distortion performance of CSVC is optimal DISCOVERfollows H264 is the worst Among the three the energyconsumption of CSVC is approximately invariant regardlessof reconstruction quality while the reconstruction quality ofH264 andDISCOVERhas a great correlationwith the energyconsumption of coding

5 Conclusions

This paper has conducted an experiment-driven analysis ofrate-distortion and rate-energy-distortion performances ofCSVC algorithm and compares them with that of H264 andDISCOVER The rate-distortion and rate-energy-distortionperformances of the three systems are evaluated under thesame experimental environment Experiment results showthat the rate-distortion performance of CSVC has a largeperformance difference from H264 and DISCOVER but itsrate-energy-distortion performance has a greater advantagethat is the rapid improvement of its reconstruction qualitydoes not depend on coding energy input Therefore onthe premise that communication bandwidth is effectivelyimproved CSVC can be used as a candidate for futurewireless video communication because of its characteristicswhich provides wireless video terminals limited by energyconsumption and computing power with more possibili-ties

At present the rate-distortion performance of CSVC isstill not ideal and there is still some way to go before weput CSVC into practical use Efforts should be made in the

International Journal of Digital Multimedia Broadcasting 7

30

35

40

45

50

55

60PS

NR

(dB)

50 100 150 200 250 300 3500Encoder time (s)

H264DISCOVERCSVC

(a) Foreman

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(b) Bus

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

60

PSN

R (d

B)

H264DISCOVERCSVC

(c) Football

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(d) Mobile

Figure 6 Comparison of rate-energy-distortion performance ofH264 DISCOVER andCSVC encoder under different test video sequences

following areas to improve the rate-distortion performanceof CSVC

(1) Side Information Estimation Rate-distortion performanceof CSVC is greatly related to the accuracy of side informationestimation which means that high-quality side informationimmensely reduces the required supply of bit load fromencoder side Therefore finding the appropriate motion esti-mation algorithm to obtain more accurate side informationand realizing the optimal reconstruction of decoding willbecome the key to improving the rate-distortion performanceof CSVC

(2) A Priori Structural Feature Modeling of Video FramesImages of the same type often have similar structuralinformation Therefore the reconstructed model can beconstructed by evaluating the structural information of thedecoded video frames and extracting the prior knowledgewhich can reduce code rate and improve the reconstructionquality For example the statistical correlation structurecan be used in the image transformation coefficient and atree structure can be adopted for the wavelet coefficientHowever due to the complexity and uncertainty of natural

images further study should be made on how to use a prioriknowledge to construct a suitable model

(3) Quantization Measurement Uniform quantization is themajor method adopted to quantify the CS measurementcurrently But the traditional entropy coding method isnot ideal for compressing the uniform quantization valuesbecause of the statistical independence between the uniformquantization values Then how to express the CS value withthe least number of bits with the constraint of information-theoretic rate-distortion coding theorem is one of the keytopics of the following research Therefore it is necessary topropose a new nonuniform quantization method to establishstatistical correlation between quantization values and herebyto design a new entropy coding method matching thestatistical correlation

The above-mentioned further researches are employedat decoder to improve the rate-distortion performance ofCSVC but the CS measuring at encoder guarantees theadvantage of CSVC in rate-energy-distortion performanceTherefore the rate-energy-distortion performance cannot beaffected while improving the rate-distortion performance ofCSVC

8 International Journal of Digital Multimedia Broadcasting

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work was supported in part by the National NaturalScience Foundation of China under Grant no 61501393in part by the Key Scientific Research Project of Collegesand Universities in Henan Province of China under Grant16A520069 in part by Youth Sustentation Fund of XinyangNormal University under Grant no 2015-QN-043 and inpart by Scientific Research Foundation of Graduate School ofXinyang Normal University under Grant no 2016KYJJ10

References

[1] J Ostermann J Bormans P List et al ldquoVideo coding withH264AVC tools performance and complexityrdquo IEEE Circuitsand Systems Magazine vol 4 no 1 pp 7ndash28 2004

[2] B Girod A M Aaron S Rane and D Rebollo-MonederoldquoDistributed video codingrdquo Proceedings of the IEEE vol 93 no1 pp 71ndash83 2005

[3] A DWyner and J Ziv ldquoThe rate-distortion function for sourcecoding with side information at the decoderrdquo IEEE Transactionson Information Theory vol 22 no 1 pp 1ndash10 1976

[4] A Aaron S Rane R Zhang and B Girod ldquoWyner-Ziv codingfor video applications to compression and error resiliencerdquo inProceedings of the Data Compression Conference (DCC rsquo03) pp93ndash102 Snowbird Utah USA March 2003

[5] R Puri A Majumdar and K Ramchandran ldquoPRISM a videocoding paradigm with motion estimation at the decoderrdquo IEEETransactions on Image Processing vol 16 no 10 pp 2436ndash24482007

[6] Q Xu and Z Xiong ldquoLayered Wyner-Ziv video codingrdquo inProceedings of the Visual Communications and Image Processing2004 pp 83ndash91 IEEE San Jose Calif USA January 2004

[7] W Liu L Dong and W Zeng ldquoMotion refinement basedprogressive side-information estimation for Wyner-Ziv videocodingrdquo IEEE Transactions on Circuits amp Systems for VideoTechnology vol 20 no 12 pp 1863ndash1875 2010

[8] X Artigas J Ascenso M Dalai S Klomp D Kubasov and MOuaret ldquoThe discover codec architecture techniques and eval-uationrdquo in Proceedings of the 26th Picture Coding Symposium(PCS rsquo07) pp 1103ndash1120 Lisbon Portugal November 2007

[9] J Slowack J Skorupa N Deligiannis P Lambert AMunteanuand R van de Walle ldquoDistributed video coding with feedbackchannel constraintsrdquo IEEE Transactions on Circuits amp Systemsfor Video Technology vol 22 no 7 pp 1014ndash1026 2012

[10] C Brites andF Pereira ldquoCorrelationnoisemodeling for efficientpixel and transform domain Wyner-Ziv video codingrdquo IEEETransactions on Circuits and Systems for Video Technology vol18 no 9 pp 1177ndash1190 2008

[11] E J Candes J Romberg and T Tao ldquoRobust uncertaintyprinciples exact signal reconstruction from highly incompletefrequency informationrdquo IEEE Transactions on InformationThe-ory vol 52 no 2 pp 489ndash509 2006

[12] D L Donoho ldquoCompressed sensingrdquo IEEE Transactions onInformation Theory vol 52 no 4 pp 1289ndash1306 2006

[13] E J Candes and M B Wakin ldquoAn introduction to compressivesampling a sensingsampling paradigm that goes against thecommon knowledge in data acquisitionrdquo IEEE Signal ProcessingMagazine vol 25 no 2 pp 21ndash30 2008

[14] R Li H Liu R Xue and Y Li ldquoCompressive-sensing-basedvideo codec by autoregressive prediction and adaptive residualrecoveryrdquo International Journal of Distributed Sensor Networksvol 2015 Article ID 562840 19 pages 2015

[15] Y Liu X Zhu L Zhang and S H Cho ldquoDistributed com-pressed video sensing in camera sensor networksrdquo InternationalJournal of Distributed Sensor Networks vol 2012 Article ID352167 10 pages 2012

[16] C Di Laura D Pajuelo andG Kemper ldquoA novel steganographytechnique for SDTV-H264AVC encoded videordquo InternationalJournal of Digital Multimedia Broadcasting vol 2016 Article ID6950592 9 pages 2016

[17] Y Lahbabi E H Ibn Elhaj and A Hammouch ldquoAdaptivestreaming of scalable videos over P2PTVrdquo International Journalof Digital Multimedia Broadcasting vol 2015 Article ID 28309710 pages 2015

[18] S Pudlewski and T Melodia ldquoCompressive video streamingdesign and rate-energy-distortion analysisrdquo IEEE Transactionson Multimedia vol 15 no 8 pp 2072ndash2086 2013

[19] D Baron M B Wakin M F Duarte S Sarvotham and R GBaraniuk ldquoDistributed compressive sensingrdquo httparxivorgabs09013403

[20] J Prades-Nebot Y Ma and T Huang ldquoDistributed videocoding using compressive samplingrdquo in Proceedings of thePicture Coding Symposium (PCS rsquo09) pp 1ndash4 Chicago Ill USAMay 2009

[21] T T Do Y Chen D T Nguyen N Nguyen L Gan and T DTran ldquoDistributed compressed video sensingrdquo in Proceedingsof the 16th IEEE International Conference on Image Processing(ICIP rsquo09) pp 1393ndash1396 IEEE Cairo Egypt November 2009

[22] E W Tramel and J E Fowler ldquoVideo compressed sensingwith multihypothesisrdquo in Proceedings of the Data CompressionConference (DCC rsquo11) pp 193ndash202 Snowbird Utah USAMarch 2011

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpswwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 7: Rate-Distortion and Rate-Energy-Distortion Evaluations of ...

International Journal of Digital Multimedia Broadcasting 7

30

35

40

45

50

55

60PS

NR

(dB)

50 100 150 200 250 300 3500Encoder time (s)

H264DISCOVERCSVC

(a) Foreman

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(b) Bus

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

60

PSN

R (d

B)

H264DISCOVERCSVC

(c) Football

50 100 150 200 250 300 350 4000Encoder time (s)

25

30

35

40

45

50

55

PSN

R (d

B)

H264DISCOVERCSVC

(d) Mobile

Figure 6 Comparison of rate-energy-distortion performance ofH264 DISCOVER andCSVC encoder under different test video sequences

following areas to improve the rate-distortion performanceof CSVC

(1) Side Information Estimation Rate-distortion performanceof CSVC is greatly related to the accuracy of side informationestimation which means that high-quality side informationimmensely reduces the required supply of bit load fromencoder side Therefore finding the appropriate motion esti-mation algorithm to obtain more accurate side informationand realizing the optimal reconstruction of decoding willbecome the key to improving the rate-distortion performanceof CSVC

(2) A Priori Structural Feature Modeling of Video FramesImages of the same type often have similar structuralinformation Therefore the reconstructed model can beconstructed by evaluating the structural information of thedecoded video frames and extracting the prior knowledgewhich can reduce code rate and improve the reconstructionquality For example the statistical correlation structurecan be used in the image transformation coefficient and atree structure can be adopted for the wavelet coefficientHowever due to the complexity and uncertainty of natural

images further study should be made on how to use a prioriknowledge to construct a suitable model

(3) Quantization Measurement Uniform quantization is themajor method adopted to quantify the CS measurementcurrently But the traditional entropy coding method isnot ideal for compressing the uniform quantization valuesbecause of the statistical independence between the uniformquantization values Then how to express the CS value withthe least number of bits with the constraint of information-theoretic rate-distortion coding theorem is one of the keytopics of the following research Therefore it is necessary topropose a new nonuniform quantization method to establishstatistical correlation between quantization values and herebyto design a new entropy coding method matching thestatistical correlation

The above-mentioned further researches are employedat decoder to improve the rate-distortion performance ofCSVC but the CS measuring at encoder guarantees theadvantage of CSVC in rate-energy-distortion performanceTherefore the rate-energy-distortion performance cannot beaffected while improving the rate-distortion performance ofCSVC

8 International Journal of Digital Multimedia Broadcasting

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work was supported in part by the National NaturalScience Foundation of China under Grant no 61501393in part by the Key Scientific Research Project of Collegesand Universities in Henan Province of China under Grant16A520069 in part by Youth Sustentation Fund of XinyangNormal University under Grant no 2015-QN-043 and inpart by Scientific Research Foundation of Graduate School ofXinyang Normal University under Grant no 2016KYJJ10

References

[1] J Ostermann J Bormans P List et al ldquoVideo coding withH264AVC tools performance and complexityrdquo IEEE Circuitsand Systems Magazine vol 4 no 1 pp 7ndash28 2004

[2] B Girod A M Aaron S Rane and D Rebollo-MonederoldquoDistributed video codingrdquo Proceedings of the IEEE vol 93 no1 pp 71ndash83 2005

[3] A DWyner and J Ziv ldquoThe rate-distortion function for sourcecoding with side information at the decoderrdquo IEEE Transactionson Information Theory vol 22 no 1 pp 1ndash10 1976

[4] A Aaron S Rane R Zhang and B Girod ldquoWyner-Ziv codingfor video applications to compression and error resiliencerdquo inProceedings of the Data Compression Conference (DCC rsquo03) pp93ndash102 Snowbird Utah USA March 2003

[5] R Puri A Majumdar and K Ramchandran ldquoPRISM a videocoding paradigm with motion estimation at the decoderrdquo IEEETransactions on Image Processing vol 16 no 10 pp 2436ndash24482007

[6] Q Xu and Z Xiong ldquoLayered Wyner-Ziv video codingrdquo inProceedings of the Visual Communications and Image Processing2004 pp 83ndash91 IEEE San Jose Calif USA January 2004

[7] W Liu L Dong and W Zeng ldquoMotion refinement basedprogressive side-information estimation for Wyner-Ziv videocodingrdquo IEEE Transactions on Circuits amp Systems for VideoTechnology vol 20 no 12 pp 1863ndash1875 2010

[8] X Artigas J Ascenso M Dalai S Klomp D Kubasov and MOuaret ldquoThe discover codec architecture techniques and eval-uationrdquo in Proceedings of the 26th Picture Coding Symposium(PCS rsquo07) pp 1103ndash1120 Lisbon Portugal November 2007

[9] J Slowack J Skorupa N Deligiannis P Lambert AMunteanuand R van de Walle ldquoDistributed video coding with feedbackchannel constraintsrdquo IEEE Transactions on Circuits amp Systemsfor Video Technology vol 22 no 7 pp 1014ndash1026 2012

[10] C Brites andF Pereira ldquoCorrelationnoisemodeling for efficientpixel and transform domain Wyner-Ziv video codingrdquo IEEETransactions on Circuits and Systems for Video Technology vol18 no 9 pp 1177ndash1190 2008

[11] E J Candes J Romberg and T Tao ldquoRobust uncertaintyprinciples exact signal reconstruction from highly incompletefrequency informationrdquo IEEE Transactions on InformationThe-ory vol 52 no 2 pp 489ndash509 2006

[12] D L Donoho ldquoCompressed sensingrdquo IEEE Transactions onInformation Theory vol 52 no 4 pp 1289ndash1306 2006

[13] E J Candes and M B Wakin ldquoAn introduction to compressivesampling a sensingsampling paradigm that goes against thecommon knowledge in data acquisitionrdquo IEEE Signal ProcessingMagazine vol 25 no 2 pp 21ndash30 2008

[14] R Li H Liu R Xue and Y Li ldquoCompressive-sensing-basedvideo codec by autoregressive prediction and adaptive residualrecoveryrdquo International Journal of Distributed Sensor Networksvol 2015 Article ID 562840 19 pages 2015

[15] Y Liu X Zhu L Zhang and S H Cho ldquoDistributed com-pressed video sensing in camera sensor networksrdquo InternationalJournal of Distributed Sensor Networks vol 2012 Article ID352167 10 pages 2012

[16] C Di Laura D Pajuelo andG Kemper ldquoA novel steganographytechnique for SDTV-H264AVC encoded videordquo InternationalJournal of Digital Multimedia Broadcasting vol 2016 Article ID6950592 9 pages 2016

[17] Y Lahbabi E H Ibn Elhaj and A Hammouch ldquoAdaptivestreaming of scalable videos over P2PTVrdquo International Journalof Digital Multimedia Broadcasting vol 2015 Article ID 28309710 pages 2015

[18] S Pudlewski and T Melodia ldquoCompressive video streamingdesign and rate-energy-distortion analysisrdquo IEEE Transactionson Multimedia vol 15 no 8 pp 2072ndash2086 2013

[19] D Baron M B Wakin M F Duarte S Sarvotham and R GBaraniuk ldquoDistributed compressive sensingrdquo httparxivorgabs09013403

[20] J Prades-Nebot Y Ma and T Huang ldquoDistributed videocoding using compressive samplingrdquo in Proceedings of thePicture Coding Symposium (PCS rsquo09) pp 1ndash4 Chicago Ill USAMay 2009

[21] T T Do Y Chen D T Nguyen N Nguyen L Gan and T DTran ldquoDistributed compressed video sensingrdquo in Proceedingsof the 16th IEEE International Conference on Image Processing(ICIP rsquo09) pp 1393ndash1396 IEEE Cairo Egypt November 2009

[22] E W Tramel and J E Fowler ldquoVideo compressed sensingwith multihypothesisrdquo in Proceedings of the Data CompressionConference (DCC rsquo11) pp 193ndash202 Snowbird Utah USAMarch 2011

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpswwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 8: Rate-Distortion and Rate-Energy-Distortion Evaluations of ...

8 International Journal of Digital Multimedia Broadcasting

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

This work was supported in part by the National NaturalScience Foundation of China under Grant no 61501393in part by the Key Scientific Research Project of Collegesand Universities in Henan Province of China under Grant16A520069 in part by Youth Sustentation Fund of XinyangNormal University under Grant no 2015-QN-043 and inpart by Scientific Research Foundation of Graduate School ofXinyang Normal University under Grant no 2016KYJJ10

References

[1] J Ostermann J Bormans P List et al ldquoVideo coding withH264AVC tools performance and complexityrdquo IEEE Circuitsand Systems Magazine vol 4 no 1 pp 7ndash28 2004

[2] B Girod A M Aaron S Rane and D Rebollo-MonederoldquoDistributed video codingrdquo Proceedings of the IEEE vol 93 no1 pp 71ndash83 2005

[3] A DWyner and J Ziv ldquoThe rate-distortion function for sourcecoding with side information at the decoderrdquo IEEE Transactionson Information Theory vol 22 no 1 pp 1ndash10 1976

[4] A Aaron S Rane R Zhang and B Girod ldquoWyner-Ziv codingfor video applications to compression and error resiliencerdquo inProceedings of the Data Compression Conference (DCC rsquo03) pp93ndash102 Snowbird Utah USA March 2003

[5] R Puri A Majumdar and K Ramchandran ldquoPRISM a videocoding paradigm with motion estimation at the decoderrdquo IEEETransactions on Image Processing vol 16 no 10 pp 2436ndash24482007

[6] Q Xu and Z Xiong ldquoLayered Wyner-Ziv video codingrdquo inProceedings of the Visual Communications and Image Processing2004 pp 83ndash91 IEEE San Jose Calif USA January 2004

[7] W Liu L Dong and W Zeng ldquoMotion refinement basedprogressive side-information estimation for Wyner-Ziv videocodingrdquo IEEE Transactions on Circuits amp Systems for VideoTechnology vol 20 no 12 pp 1863ndash1875 2010

[8] X Artigas J Ascenso M Dalai S Klomp D Kubasov and MOuaret ldquoThe discover codec architecture techniques and eval-uationrdquo in Proceedings of the 26th Picture Coding Symposium(PCS rsquo07) pp 1103ndash1120 Lisbon Portugal November 2007

[9] J Slowack J Skorupa N Deligiannis P Lambert AMunteanuand R van de Walle ldquoDistributed video coding with feedbackchannel constraintsrdquo IEEE Transactions on Circuits amp Systemsfor Video Technology vol 22 no 7 pp 1014ndash1026 2012

[10] C Brites andF Pereira ldquoCorrelationnoisemodeling for efficientpixel and transform domain Wyner-Ziv video codingrdquo IEEETransactions on Circuits and Systems for Video Technology vol18 no 9 pp 1177ndash1190 2008

[11] E J Candes J Romberg and T Tao ldquoRobust uncertaintyprinciples exact signal reconstruction from highly incompletefrequency informationrdquo IEEE Transactions on InformationThe-ory vol 52 no 2 pp 489ndash509 2006

[12] D L Donoho ldquoCompressed sensingrdquo IEEE Transactions onInformation Theory vol 52 no 4 pp 1289ndash1306 2006

[13] E J Candes and M B Wakin ldquoAn introduction to compressivesampling a sensingsampling paradigm that goes against thecommon knowledge in data acquisitionrdquo IEEE Signal ProcessingMagazine vol 25 no 2 pp 21ndash30 2008

[14] R Li H Liu R Xue and Y Li ldquoCompressive-sensing-basedvideo codec by autoregressive prediction and adaptive residualrecoveryrdquo International Journal of Distributed Sensor Networksvol 2015 Article ID 562840 19 pages 2015

[15] Y Liu X Zhu L Zhang and S H Cho ldquoDistributed com-pressed video sensing in camera sensor networksrdquo InternationalJournal of Distributed Sensor Networks vol 2012 Article ID352167 10 pages 2012

[16] C Di Laura D Pajuelo andG Kemper ldquoA novel steganographytechnique for SDTV-H264AVC encoded videordquo InternationalJournal of Digital Multimedia Broadcasting vol 2016 Article ID6950592 9 pages 2016

[17] Y Lahbabi E H Ibn Elhaj and A Hammouch ldquoAdaptivestreaming of scalable videos over P2PTVrdquo International Journalof Digital Multimedia Broadcasting vol 2015 Article ID 28309710 pages 2015

[18] S Pudlewski and T Melodia ldquoCompressive video streamingdesign and rate-energy-distortion analysisrdquo IEEE Transactionson Multimedia vol 15 no 8 pp 2072ndash2086 2013

[19] D Baron M B Wakin M F Duarte S Sarvotham and R GBaraniuk ldquoDistributed compressive sensingrdquo httparxivorgabs09013403

[20] J Prades-Nebot Y Ma and T Huang ldquoDistributed videocoding using compressive samplingrdquo in Proceedings of thePicture Coding Symposium (PCS rsquo09) pp 1ndash4 Chicago Ill USAMay 2009

[21] T T Do Y Chen D T Nguyen N Nguyen L Gan and T DTran ldquoDistributed compressed video sensingrdquo in Proceedingsof the 16th IEEE International Conference on Image Processing(ICIP rsquo09) pp 1393ndash1396 IEEE Cairo Egypt November 2009

[22] E W Tramel and J E Fowler ldquoVideo compressed sensingwith multihypothesisrdquo in Proceedings of the Data CompressionConference (DCC rsquo11) pp 193ndash202 Snowbird Utah USAMarch 2011

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpswwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of

Page 9: Rate-Distortion and Rate-Energy-Distortion Evaluations of ...

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpswwwhindawicom

VLSI Design

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation httpwwwhindawicom

Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

DistributedSensor Networks

International Journal of


Top Related