变比特率均值匹配相关矢量量化图像编码算法

矢量量化(VQ)是一种有效的数据压缩技术。为找出与输入矢量最匹配的码字,传统的穷尽搜索矢量量化编码算法需要计算输入矢量与所有码字之间的失真测度。码书大小和矢量维数越大,穷尽搜索矢量量化编码的计算复杂度就越高。为了降低穷尽搜索矢量量化器的编码复杂度,本文提出了一种用于快速图像编码的均值匹配相关矢量量化器(MMCVQ)。在编码前,首先计算所有码字的均值,然后按照这些均值从小到大对码书进行排序。编码阶段,利用邻近图像块的高度相关性和当前输入矢量的均值共同确定相应的码字搜索范围。实验结果表明,当阈值大小为320时,与传统穷尽搜索矢量量化编码法相比,虽然MMCVQ算法的编码质量下降约0.3～0.4dB,但速度快14倍而且比特率下降0.1～0.2比特像素。

An Image Encoding Algorithm Based on Variable-Rate Meanmatch Correlation Vector Quantization

Vector quantization(VQ)is an efficient data compression technique.To find the nearest codeword for each input vector,traditional full-search vector quantization encoding al-gorithm requires calculating the distance between each codeword and the input vector.The larger the codebook size or the vector dimension is,the higher the computational complexity of the full-search VQ is.In order to reduce the computational load of the full-search VQ,a mean-match correlation vector quantizer,MMCVQ,is presented for fast image encoding in this paper.Before encoding,the mean values of all codewords are computed and a sorted codebook is obtained according to the ascending order of the mean values of codewords.Dur-ing the encoding stage,high correlation of the adjacent image blocks and the mean value of the input vector are utilized to determine the searching range.The experimental result shows that,compared with the conventional full search VQ encoding algorithm,although the en-coding quality degradation of the MMCVQ encoding algorithm is about 0.3-0.4dB,the MMCVQ can accelerate the encoding speed by14times and can reduce the encoding bit rate by0.1-0.2bits?pixle,when the threshold value is set to be320.