Perceptual dithering for octave subband image coding

Abstract

Octave subband coding has been used as an efficient image coding approach for various well-known standards such as MPEG-4 and JPEG-2000. Although the sub-images in the octave subband representation have low statistical correlation, the visual similarity among the sub-images indicates that there exists some unexploited perceptual redundancy between sibling bands in addition to the parent-child redundancy. In this paper, we show that the original image can be perceptually dithered to form a visually equivalent image with increased interband correlation that can be used to achieve more compression. To remove the redundancy among the sibling subbands of an image, we provide a novel perceptual dithering approach that is based on entropy reduction technique with perceptual model. The theoretical basis for the entropy reduction is proven by a theorem for Gaussian distributed signals. The use of our dithering technique is demonstrated for an MPEG-4 compliant encoder with improved coding efficiency using MPEG-4 Visual Texture Coding (VTC) based on zerotree entropy coding. Our results show that there exists a perceptual interband redundancy even though the original interband correlation is small. For a perceptually transparent image quality, our approach can achieve bit savings over MPEG-4 VTC by 11-30%, while maintaining compatibility with the MPEG-4 standards. (C) 2003 Elsevier Inc. All rights reserved.