Perceptual Mpeg-2 Syntactic Information Coding: a Performance Study Based on Psychophysics

This work addresses the optimization of TVresolution MPEG-2 video streams to be transmitted over packet networks facing a given packet loss probability. Several studies have already been carried out on this speci c problem [1, 2, 3]. The main contribution of this work resides in the design of a new exible scene-complexity adaptive mechanism, namely the Perceptual Syntactic Information Coding (PSIC) mechanism. The PSIC adaptively modulates the number of macroblocks per slice in order to optimize the global video quality which depends on both the video encoding quality and the video degradations due to data loss during the transmission. We compare its performance against a xed-length slice scheme to be transmitted over ATM networks using AAL-5. All the quality assessments are based on a perceptual quality metric, MPQM 1, which proved to behave consistently with human judgements [4]. Results show that the proposed mechanism behaves much better in terms of global perceptual video quality. Published in Picture Coding Symposium, Berlin, September 1997