Perceptual rate control algorithms for fax-based video compression

Abstract

Summary form only given. Video samples usually are predicted from coded versions of nearby samples sent either earlier in the same frame or in the previous frame. Analysis of the human vision system (HVS) suggests that we may not need to correct values of residuals that do not exceed a perceptual threshold sometimes referred to in the literature of perception as the just-noticeable-distortion (JND). The ideal JND provides each pixel being coded with a threshold level below which discrepancies are perceptually distortion-free. Also of interest is the rate control analysis of noticeable, above threshold distortions that inevitably result at low bit rates. Because facsimile-based video compression (FBVC) processing is done in the spatio-temporal pixel domain, we can exploit the local psycho-perceptual properties of the HVS. Our proposed rate control algorithms are distinguished by being computationally economical, transform-free, devoid of block-based artifacts, and capable of easily providing a constant bit rate video stream.