The Frequency-Domain Effects of Stochastic Image Foveation in Superpixelating CamerasReview – Motivation and ObjectiveNontrivial Translation of Control SignalMy Custom Methods – vasiHalftone, vasiHalftone2F&S Error DiffusionThe RestConclusionsReferencesThe Frequency-Domain The Frequency-Domain Effects of Stochastic Image Effects of Stochastic Image Foveation in Superpixelating Foveation in Superpixelating CamerasCamerasThayne CoffmanThayne CoffmanEE381K-14EE381K-14May 3, 2005May 3, 2005Review – Motivation and ObjectiveReview – Motivation and ObjectiveSuperpixellating (“VASI”) cameras Superpixellating (“VASI”) cameras Translation of desired resolution to Translation of desired resolution to control signal can be done by control signal can be done by halftoninghalftoningWhich halftoning method will give the Which halftoning method will give the best ATR performance?best ATR performance?Block error diffusionBlock error diffusionBlue noise ditheringBlue noise ditheringFloyd & Steinberg error diffusionFloyd & Steinberg error diffusionRaster scanRaster scanSerpentine scanSerpentine scanClassical screeningClassical screening9-level clustered dot9-level clustered dot9-level dispersed dot9-level dispersed dotWhite noiseWhite noiseA method of my own designA method of my own designvasiHalftone, vasiHalftone2vasiHalftone, vasiHalftone2Not exactly halftoning algorithmsNot exactly halftoning algorithms[McCarley et al, 2004][McCarley et al, 2004]Performance measured byPerformance measured byPSNRPSNRWSNRWSNRLDM (not a useful LDM (not a useful differentiator)differentiator)UQIUQINontrivial Translation of Control Nontrivial Translation of Control SignalSignalControl signal vs. realized bandwidthControl signal vs. realized bandwidthNontrivial relationship caused by the geometry of pixel sharing patternsNontrivial relationship caused by the geometry of pixel sharing patternsRequires customization (inverse function) of control signal for each Requires customization (inverse function) of control signal for each halftoning methodhalftoning methodStairstep patterns limit your control over actual realized bandwidthStairstep patterns limit your control over actual realized bandwidthDetails in paperDetails in paperMy Custom Methods – vasiHalftone, My Custom Methods – vasiHalftone, vasiHalftone2vasiHalftone2Semi-regularly spaced rectangles, size depends on desired bandwidthSemi-regularly spaced rectangles, size depends on desired bandwidthFor a given control signalFor a given control signalConsistently superior PSNR & WSNRConsistently superior PSNR & WSNRConsistently overshot desired bandwidth by ~30-100%Consistently overshot desired bandwidth by ~30-100%They were essentially cheating by using extra bandwidthThey were essentially cheating by using extra bandwidthAs currently designed, these methods have very poor bandwidth controlAs currently designed, these methods have very poor bandwidth controlOriginalOriginalSharing SignalSharing SignalResulting ImageResulting ImagePSNR = 13.3 dBPSNR = 13.3 dBWSNR = 16.9 WSNR = 16.9 dBdBDesired BW = Desired BW = 9.6%9.6%Actual BW = Actual BW = 18.8%18.8%Inflation = 97%Inflation = 97%F&S Error DiffusionF&S Error DiffusionGood performance and good bandwidth controlGood performance and good bandwidth controlGood SNR in ROIs means accurate ATRGood SNR in ROIs means accurate ATRGood SNR in non-ROIs means good target acquisitionGood SNR in non-ROIs means good target acquisitionGood bandwidth control means precise VASI frame rate Good bandwidth control means precise VASI frame rate controlcontrolOriginalOriginalSharing SignalSharing SignalResulting ImageResulting ImagePSNR = 17.5 dB (33.3 dB in ROI)PSNR = 17.5 dB (33.3 dB in ROI)WSNR = 16.4 dB (33.8 dB in ROI)WSNR = 16.4 dB (33.8 dB in ROI)Desired BW = 11.6%Desired BW = 11.6%Actual BW = 12.1%Actual BW = 12.1%Inflation = 4%Inflation = 4%The RestThe RestMethodMethodPerformance Performance (SNR)(SNR)Bandwidth controlBandwidth controlBlock error diffusionBlock error diffusionPoorPoorGoodGoodClassical screeningClassical screeningDecentDecentPoorPoorStochastic methodsStochastic methodsPoorPoorSuceptible to Suceptible to “catastrophic gray-“catastrophic gray-out”out”OriginalOriginalBlue noiseBlue noiseBlock error diffusionBlock error diffusionOriginalOriginalClustered dotClustered dotDispersed dotDispersed dotWhite noiseWhite noiseConclusionsConclusionsFloyd & Steinberg error diffusion gives the best results Floyd & Steinberg error diffusion gives the best results while still being able to control bandwidth preciselywhile still being able to control bandwidth preciselyvasiHalftone & vasiHalftone2 vasiHalftone & vasiHalftone2 Consistently the best PSNR, WSNRConsistently the best PSNR, WSNRPoor bandwidth control – overshot specifications by 30-100%Poor bandwidth control – overshot specifications by 30-100%Bandwidth inflation means it’s not a fair comparison (they’re Bandwidth inflation means it’s not a fair comparison (they’re cheating)cheating)Stochastic methods (white & blue noise) perform poorlyStochastic methods (white & blue noise) perform poorlyOutperformed by deterministic approachesOutperformed by deterministic approachesSusceptible to “catastrophic gray-out”Susceptible to “catastrophic gray-out”Classical screening performs marginally Classical screening performs marginally andand has bad bandwidth controlhas bad bandwidth controlReferencesReferencesP. McCarley, M. Massie, J.P. Curzan, “Large format P. McCarley, M. Massie, J.P. Curzan, “Large format variable spatial acuity superpixel imaging: visible and variable spatial acuity superpixel imaging: visible and infrared systems applications,” infrared systems applications,” Proc. SPIE, Infrared Proc. SPIE, Infrared Technology and Applications XXXTechnology and Applications XXX[sic.], vol. 5406, pp. [sic.], vol. 5406, pp. 361-369, Aug 2004.361-369, Aug 2004.V. Monga, N. Damera-Venkata, B. Evans, V. Monga, N. Damera-Venkata, B. Evans, Halftoning Toolbox for MatlabHalftoning Toolbox for Matlab.. Version 1.1 released Version 1.1 released November 7, 2002. Available online at November 7, 2002. Available online at http://http://www.ece.utexas.edu/~bevans/projects/halftoningwww.ece.utexas.edu/~bevans/projects/halftoning//.
View Full Document
Unlocking...