We seek to marry rate-distortion theory with perceptual science in order to develop improvements in fidelity criteria and techniques for the compression of video and other natural data sources. The information theory of Weber-Fechner (WF) and root law (RL) fidelity criteria will be studied, with emphasis on their application to lossy compression of video signals and on calculating lower bounds to their rate-distortion functions. Multiterminal source coding models will be employed to deal with perceptual considerations, such as just-noticeable-differences and masking, that arise when sequences of video frames are viewed. Also, we shall explore novel video compression schemes grounded in WF/RL theory that provide dramatic computational simplifications relative to competing algorithms at high data rates and promise better quality in the low-bit-rate regime for such appications as video telephony and distance learning. Low-latency video compression and transport will be performed on facilities in the DISCOVER Lab to test our perceptually based algorithms.
