National Imagery Interpretability Rating Scalesdeveloped in 1970s by Imagery Resolution and Reporting Standards (IRARS)as a mean to define requirements for Intelligence, Surveillance and Reconnaissance systems 10-level rating scale (0 - 9) indicate the interpretability of an imageeach level defines objects that should be able to be observed within imagesis to be estimated by human operatorsNATO Imagery Interpretability Rating Scales (STANAG 7194)

NIIRS LEVEL 4 NIIRS LEVEL 7

identify farm buildings: barns, silos, residences.detect basketball court, tennis court.identify individual tracks, rail pairs.identify control towersdetect individual steps on a stairway.identify individual railroad ties.detect stumps and rocks in forest clearings and meadow.

GENERAL IMAGE QUALITY EQUATIONproposed by Leachtenauer in 1997mathematical estimation of NIIRS

GSDGM - the geometric mean of the ground sampled distanceRERGM - the geometric mean of the relative edge response HGM - the geometric mean of height overshoot G - the noise gain SNR - the signal to noise ratioIf the RER exceeds 0.9, then a equals 3.32 and b equals 1.559; otherwise, a equals 3.16 and b equals 2.817

GROUND SAMPLE DISTANCE (GSD)projection of the pixel to the groundthe most important parameter on GIQEmeasured in inches

p - the pixel pitchf - focal length of the sensorh - the altitude of the platform - the look angle

It is measured in X and Y direction:the result is the geometric mean

EDGE RESPONSEEdge Spread Functionrepresent the edge profile across an edgeslanted edge method describe in ISO 12233

ideal slanted edge ESF with 32 pixels sampled distance

EDGE RESPONSE

slanted edgeblurred slanted edge

high respone low response

SLANTED EDGE METHOD

SLANTED EDGE METHOD

best vertical edge

best horizontal edge Image to be processed

RELATIVE EDGE RESPONSE (RER)the second largest contributing factor to the GIQErepresents the slope of the normalized edge responseis measured between two points one-half pixel from an edgethe edge response is normalized to the range of 0 to 1it is measured in the x and y axis and the geometric mean computed

Edge Overshoot (H)height of overshoot across an edgeit is measured in the x and y axis and the geometric mean computed

high overshoot edge

Edge profile

NOISE GAIN (G) AND SIGNAL TO NOISE RATIO(SNR)G is the increase in noise resulting from MTFC (modulation transfer function compensation)SNR quantify how much a signal has been corrupted bynoisewe used constant value published in the literature (Ryan et al., 2003) for G = 4.16 and SNR = 10

NIIRS= 6.35 GSD = 0.322, RER = 0.87, H = 0.16, G = 4.16, SNR = 10

Influence of SNR in NIIRSInfluence of G in NIIRS

SNRPercentNIIRS100%6.3519729.9-10.1 10%6.350526-6.3533897.5-12.5 25%6.30427-6.3805935-15 50%6.208868-6.3996730.22033-17.7967 77.9%0 -6.4146650.0-20 100%-1.4*en-6.423524, n->

GPercentNIIRS4.160%6.3519723.744-4.576 10%6.366282-6.3376613.12-5.2 25%6.387748-6.3161962.08-6.24 50%6.423524-6.280420-8.32 100%6.495076-6.208868

SOFTWAREprogramming stuff:JavaNetbeans

NC3A applications:NSIF Viewer (NATO Secondary Imagery Format STANAG 4545 )NIMMOI (NATO Intelligence, Surveillance and Reconnaissance Metadata and Motion Imagery STANAG 4609)

open source applications: Gavaghan Geodesy LibraryImageJ

TESTING ON STILL IMAGERY

TESTING ON STILL IMAGERY

296 NIIRS images graphic

TESTING ON STILL IMAGERYCHALLENGESbig image size couldnt be processedvery long time in processing time processing increase very fast with the number of pixelsthe algorithm couldnt be applied to poor image (in edges)

SOLUTIONSsplit big images in less images sizemake NIIRS image map and color each region with it NIIRS value

TESTING ON STILL IMAGERY

Normal size

NIIRS map

Big image size (zoomed out 7X)

TESTING ON MOTION IMAGERYCHALLENGES:frames must be processed in real timeSOLUTIONSusing multithreading technique

QUESTIONS

ERROR IN PROCESSING

HIGH NIIRS

NEGATIVE NIIRS

*