Adaptation of vision model based distortion metrics to perceptual image coding

Recent research efforts in the development of objective quality and impairment measures for quality assessment of digital video and television have helped to improve and refine models for the human visual system. In particular, recent publications by S. Winkler (see "Vision Models and Quality Metrics for Images Processing Applications", PhD thesis, EPFL, Lausanne, 2000), J. Lubin and D. Fibush, (see T1A1.5 Working Group Document #97-612, ANSI T1 Standards Committee, 1997) and Z. Yu et al. (see Proc. IEEE, Jan. 2002) have incorporated comprehensive vision models as part of their picture quality and impairment metric design. Applications of these vision-model-based perceptual quality or impairment metrics to image/video compression are seen as the next step in delivering perceptual image or video coding systems which cater for improvement in perceived picture quality. This paper subscribes to an approach to perceptual image coder design whereby a vision-model-based perceptual distortion metric is introduced into a conventional coder, e.g. JPEG2000 compliant coder, in place of the mean-square-error measure for rate-distortion optimization. Simulation results show that the new perceptual image coder provides better performance over the JPEG2000 coder with or without the visual masking option (see Taubman, D., IEEE Trans. Image Proc., vol.7, p.1158-70, 2000).