Context-based and perceptual-based wavelet image compression with application to JPEG2000

Compression of digital images has been a research topic for many years. A number of successful compression techniques have been developed. Some of these techniques have evolved into international standards such as JPEG and JPEG2000. JPEG2000 is the new international standard for image compression. Apart from offering superior compression performance, it also has very useful features such as resolution scalability and quality scalability.; This dissertation focuses on some key technologies of JPEG2000, namely context-based arithmetic coding and rate-distortion optimization. First, the contexts used in JPEG2000 are analyzed using the mutual information, which is closely related to the compression performance. When combining the contexts, the mutual information between the contexts and the encoded data will decrease unless the conditional probability distributions of the combined contexts are the same. Given a desired number of contexts, an optimization procedure is proposed to determine the set of contexts that would maximize the mutual information. Experimental results show that the JPEG2000 contexts capture very well the correlations among the wavelet coefficients. At the same time, the number of contexts used as part of the standard can be reduced without loss in the coding performance.; Second, a distortion-controlled JPEG2000 encoding method is proposed to reach consistent reconstructed image quality while achieving the lowest bit rate. For this purpose, a vision model that incorporates various masking effects of human visual perception and a perceptual distortion metric that takes spatial and spectral summation of individual quantization errors into account are incorporated within the JPEG2000 standard. Compared with the conventional rate-based JPEG2000 encoding; the new method provides a way to generate images with consistent quality at a lower bit rate.; In addition, independent of the JPEG2000 effort and prior to its finalization, an embedded zerotree wavelet image coder based on intraband partitioning (EZIP) is developed and is presented here for comparison. Experimental results show that comparable performance to JPEG2000 can be achieved at a much reduced complexity. Based on randomization and data partitioning in the wavelet transformed domain, the EZIP coder is extended to alleviate the impact of channel errors on the reconstructed image quality and is incorporated into an end-to-end video coding and transmission system.