Embedded Image Coding Based on Context Classification and Quadtree Ordering in Wavelet Packet Domain

This paper presents an embedded wavelet packet image coding algorithm which is based on context classification and quadtree ordering (CCAQO). To match up well with the quadtree-based embedded coder, a new cost function for the best basis selection is adopted in discrete wavelet packet transform (DWPT). In addition, combined with the structure of the best wavelet packet basis, a complete quadtree representation of wavelet packet coefficients is established to explore the ordering procedure of embedded image coding. Since subband correlation is mainly captured by significance coding, our focus is on significance coding. The significance probability of wavelet coefficient is estimated by convoluting a 9×9 FIR filter matrix kernel with the significance states of neighboring coefficients, and then a context classifier based on LloydMax algorithm is used to categorize the wavelet coefficients with the same or similar significance probability into several contexts. The significance state of wavelet coefficient with respect to a given threshold is encoded using adaptive arithmetic coder based on the classified context. Due to the optimal context classifier and the flexible quadtree representation ability of wavelet packet coefficients, the proposed CCAQO embedded image coder offers improvement in subjective and objective quality for texture-rich images and experimental results show that it offers coding performance superior to or comparable to the state-of-the-art image coders.