Research On Embedded Image And Video Coding Based On Lifting Scheme

The development of Network technology gives rise to new requirement on image and video coding. The embedded coding algorithm based on wavelet transform, characterizes as scalable in resolution and quality (from lossly transmission to lossless transmission ), a strong error-resilience performance, Low memory requirements and rapid encoding and decoding and many other remarkable advantages, is regarded as hotspot in the research field of present image compression .The detailed analysis of the two main algorithms is present, namely, embedded zerotree wavelet coding algorithm and coding algorithms based on embedded block. Under the circumstance of low bit-rate, the algorithms referred are improved and extends to three-dimensional wavelet video coding. The main research presents as the following:Firstly, the paper puts forward line-based integer wavelet image compression algorithm. Resorting integer transformation to complete lifting scheme, it reads into the corresponding number of lines accordance to the level of decomposition and length of filter and uses internal storage repeatedly. Based on line-based wavelet transform, improved scheme of binary arithmetic coder is present. Aiming at the feature of line-base wavelet, context model is constructed. While reducing the memory requirement of wavelet image coding greatly, this algorithm can achieve acceptable compression efficiency and image quality.Secondly, the paper presents ROI coding algorithm based on Optimizing rate allocation. According to principles of optimization and PCRD-opt (post-compression rate-distortion optimization), estimations of distortion of code-blocks that contribute to ROI are enlarged through factors of interest. As it comes to the code blocks that get comparatively more allocation of rate, the priority of ROI region is accordingly enhanced and the coefficient in ROI region get priority in coding and decoding and simultaneously maintains a certain background effect.Thirdly, the paper brings out improved Visual SPIHT algorithms. From the condition of big threshold the performance of SPIHT is analyzed and low frequency subband is separately handled. According to the size of binary wavelet, low frequency subband is decomposed virtually so as to achieve longer zero trees. When VSPIHT works at middle or high bit-rates, its performance will decline. For the sake of increase efficiency, this chapter raised two location-judging functions, which identify...