Rate-perceptual-distortion optimization (RpDO) based picture coding - issues and challenges

Perceptual image and video coders have been reported extensively in the literature with a recent survey listing sixteen representative PVCs (perceptual video coders) for single-view video coding, ten of which conform to H.263, MPEG-4 or H.264/AVC international standard. Successful approaches to perceptual picture coding are classified as perceptual predictive coding, perceptual quantization, rate-perceptual-distortion optimization (RpDO) and perception-based pre-, loop- and postfiltering. This contribution analyzes and examines the performance and tradeoffs of representative perceptual as well as traditional waveform based picture coding methods in terms of their effectiveness in quality constrained coding and RpDO criterion in the context of visual quality controlled or regulated visual communication, broadcasting, entertainment and consumer electronic products, systems and services or applications. Fundamental flaws of performance benchmarking using a rate-driven coding strategy are interrogated for visual quality assured services or user experience. It is revealed that while existing perceptual distortion measures (PDMs) successfully grade distortions corresponding to HVS (human visual system) perception, their quantitative correspondence, or lack of it, to discernible levels by human visual perception in terms of just-not-noticeable-difference (JNND), just-noticeabledifference level 1 (JND1), JND level 2 (JND2), etc. has not been thoroughly investigated nor established, and this motivates the related initial work reported in this paper. Issues and constraints of existing picture coding standards are discussed in the context of developing a future picture coding standard framework for constant visual picture quality systems and services.