基于自适应搜索范围及顺序的快速矢量量化编码算法

本文提出了一种新颖的快速矢量量化编码算法.该算法在编码前预先计算每个码字的四个特征量,然后根据各特征量的升序排列分别对码字进行排序以生成四个排序码书.在编码过程中,对于不同的输入矢量,自适应产生不同的动态码字搜索范围及顺序而排除大部分码字.测试结果表明,本文算法只需搜索3%到8%码字而获得与穷尽搜索算法相近的编码质量,实际编码时间减少约93%.

A Fast Vector Quantization Encoding Algorithm Based on Adaptive Searching Range and Sequence

A novel fast encoding algorithm is presented for vector quantization.Before encoding,four characteristic values are first computed for each codeword,then four ordered codebooks are obtained according to the ascending order of four characteristic values respectively.During the encoding process,the adaptive searching range and sequence are obtained for each input vector in order to reject a lot of unmatched codewords.The test result shows that this algorithm enables us to search merely 3% to 8% of the codewords to encode an input vector while preserving its coding performance very close to that of the full search method,and the reduction of encoding time is about 93% against full search method.