+ All Categories
Home > Documents > About the Author: Witold Pedrycz Department of Electrical and Computer Engineering, University of...

About the Author: Witold Pedrycz Department of Electrical and Computer Engineering, University of...

Date post: 19-Dec-2015
Category:
Upload: ada-montgomery
View: 221 times
Download: 5 times
Share this document with a friend
50
Local descriptors and similarity measures for frontal face recognition: A comparative analysis 小小小小 小小小 小小小 小小小
Transcript
  • Slide 1
  • Slide 2
  • Slide 3
  • About the Author: Witold Pedrycz Department of Electrical and Computer Engineering, University of Alberta, Canada Professor & Canada Research Chair & IEEE Fellow & Professional Engineer Research interests and activities: Software Engineering System modelling and knowledge discovery Reconfigurable and evolvable architectures. Pattern recognition Personal Homepage http://www.ece.ualberta.ca/~pedrycz/index.html
  • Slide 4
  • About the Author: Marek Reformat A member of the IEEE and ACM. A member of program committees of several conferences related to computational intelligence and software engineering. Actively involved in North American Fuzzy Information Processing Society (NAFIPS). Research interests and activities: Knowledge extraction and knowledge representation Semantic-based intelligent systems Decision support Software quality and maintenance Personal Homepage http://www.ece.ualberta.ca/~reform/index.html
  • Slide 5
  • Contents Main process Local descriptors Gabor filter with LBP Similarity measures & Experimental results
  • Slide 6
  • Taxonomy On pixel LBP,CS,RI Psychological WLD Ternary LTP,DLTP Distance based TPLBP,FPLBP Local descriptors Rotation invar Shifted LBP Selete subset U2,DLBP,SELBP On averaged ILBP,MBLBP,GR AB Three Dimen VLBP,LBP-TOP MultiresolutionGa bor,MB,GRAB local descriptors on Gabor filtered image Gabor magnitude LGBP,MHLVP,LGBPHS,MU LGBP Gabor magnitude & phase ELGBP,MBP 3D GV- LBPTOP Gabor phase quantization HGPP,LGPDP,LGXP Derivative patterns ELGBP,LDP
  • Slide 7
  • Main process 1 2 3
  • Slide 8
  • Main process Local pattern: pixel level description
  • Slide 9
  • Contents Main process Local descriptors Gabor filter with LBP Similarity measures & Experimental results
  • Slide 10
  • Local binary patterns: Circular LBP Bilinear interpolation of a pixel Basic LBP 57 Uniform-Ri LBP B A Elongated LBP New Variants Dominant LBP Statistically Effective LBP Hamming LBP Baisc LBP etc
  • Slide 11
  • Local binary patterns: Threshold 100.11 ILBP
  • Slide 12
  • Local binary patterns: Threshold L2 L3L4 Magnitude L1 Sign ELBP
  • Slide 13
  • Local binary patterns: Magnitude Sign Original image Center gray level Local difference M S clbp_S clbp_M clbp_C clbp_map classifier clbp_Histogram CLBP
  • Slide 14
  • Local binary patterns: DLTP=| LTPU-LTPL|=135-40=95 LTP 5 (AELTP) LTP
  • Slide 15
  • Local binary patterns: Soft-LBP
  • Slide 16
  • Gabor filter with LBP: SILT P [64(1-t) 64(1+t)] t=0.1
  • Slide 17
  • Local binary patterns: Integral image D=4+1-(2+3) MB-LBP
  • Slide 18
  • Local binary patterns: GRAB(General Region Assigned to binary) 5 GARB # noise & variations & rotation tolerant operator solving the orientation problem small variation in edge angles cause smaller variations in the binary representation
  • Slide 19
  • Local binary patterns: CS-LBP
  • Slide 20
  • Local binary patterns: TP-LBP
  • Slide 21
  • Local binary patterns: FP-LBP
  • Slide 22
  • Local binary patterns: LDP 1 Robust against Gaussian white noise and non-monotonic illumination changes 2 Rotation invariant
  • Slide 23
  • Local binary patterns: VLBP
  • Slide 24
  • Local binary patterns: VLBP
  • Slide 25
  • Local binary patterns: LBP-TOP
  • Slide 26
  • Local binary patterns: LBP-TOP
  • Slide 27
  • Local binary patterns: LBP-TOP
  • Slide 28
  • Local binary patterns: LBP-TOP
  • Slide 29
  • Contents Main process Local descriptors Gabor filter with LBP Similarity measures & Experimental results
  • Slide 30
  • Gabor filter with LBP
  • Slide 31
  • Gabor filter with LBP
  • Slide 32
  • Slide 33
  • = = + +
  • Slide 34
  • Gabor filter with LBP Gabor Dennis Gabor, 1946
  • Slide 35
  • Slide 36
  • Slide 37
  • Slide 38
  • Slide 39
  • Gabor filter with LBP: 10 GGPP 80 GGPP HGP P
  • Slide 40
  • Gabor filter with LBP: GGPP LGPP HGP P
  • Slide 41
  • Gabor filter with LBP: LGPD P
  • Slide 42
  • Gabor filter with LBP: LGXP
  • Slide 43
  • Contents Main process Local descriptors Gabor filter with LBP Similarity measures & Experimental results
  • Slide 44
  • Similarity measures
  • Slide 45
  • Fusing sub-region 1: All sub-regional histograms concatenated 2: Sub-regions of two images are compared pair-wise and the results are aggregated
  • Slide 46
  • Fusing sub-region I2I2 I1I1 I1I1 I2I2 LGX P LGBP_mag A:Feature level fusion
  • Slide 47
  • Fusing sub-region LGBP_mag I1I1 I2I2 I2I2 I1I1 LGXP Others measure: 1:cosine distance measure 2:LDA 1: AdaBoost 2: Borda count Reduction dimensionality
  • Slide 48
  • Slide 49
  • Slide 50
  • Slide 51
  • Thank you

Recommended