Color image compression by stack-run-end coding

Summary form only given. We present a new wavelet based image coding algorithm for color image compression. The key renovation of this algorithm is based on a new context oriented information conversion for data compression. A small number of symbol sets were then designed to convert the information from the wavelet transform domain into a compact data structure for each subband. Unlike zerotree coding or its variations which utilize the intersubband relationship into its own data representation where hierarchical or parents-children dependency is performed, our work is a low complexity intrasubband based coding method which only addresses the information within the subband or combines the information across the subbands. The scheme works first by color space conversion, followed by uniform scalar quantization. A concise data structure which categorizes the quantized coefficients into (stack, run, end) data format is performed, where the raster scanning order for individual subband is the most common used method but predefined scanning order will also work. Compared with the standard stack-run coding, our method generalized the symbol representation and the extension of the symbol alphabets. The termination symbols which carry the zero value information towards the end of the subband or across the subbands till the end of the image help to speed up the decoding processes. Our experiment results show that our approach is very competitive to the refinement of zerotree type schemes.