Lossless Compression Schemes of Vector Quantization Indices Using State Codebook

摘要

In a memoryless vector quantization system, eachimage block is independently encoded as a correspondingindex and then an index table will be generated. In thispaper, we propose two novel schemes to compress the indextable without introducing extra encoding distortion. Bothour schemes exploit the characteristic that the blocks ofimages are highly correlated to find the same or similarindex from the neighboring indices in the compressionprocess. To increase the compression, the principalcomponent analysis (PCA) technique is also employed tosort the codewords in the codebook for minimizing thedifference of neighboring indices. In addition, our secondscheme excludes the repetitive indices from the search pathto further decrease the bit rate. Simulation results showthat our schemes are superior to SOC and traditionalmemoryless VQ on the compression rate.