基于图像块动态调整的码字内再匹配矢量量化

Vector quantization based on dynamic image block regularization and rematching in the codeword

提出一种基于图像块动态调整的码字内再匹配矢量量化方法。该方法在编码前,首先分析待编码子图像与其八邻域子图像的相似性,通过给定的阈值判断是否相似,如果相似,则用同一个码字编码;否则就单独编码。在编码时,由于匹配码字只是和子图像整体上失真度最小,所以进一步把子图像和匹配码字划分为小块,然后子图像中的每一小块在匹配码字中再匹配。实验结果表明,相对于普通矢量量化,该方法不但可以提高编码速度,而且图像质量也有明显改善。

An method of vector quantization based on dynamic image block regularization and rematching in the codeword was presented. Before encoding, the similarity of the encoding image block and its eight-neighbour blocks was analyzed, if the similarity value is more than the threshold, the neighbour block and the encoding image block would use the same codeword to encode, otherwise, the neighbour block would encode singly. When encoding, since the distrotion only represented the global deviation between the matching codeword and the encoding image block, they were divided into small blocks, and the small image blocks were rematched in the matching codeword. Experiment results show that this method can improve the encoding speed and image restoration performance against the normal VQ.