矢量量化图像压缩方法

Image Compression Method Based on Vector Quantization

摘要:

实现并讨论一种数字图像压缩编码方法。该方法按照LBG算法以均方误差法计算两向量之间的距离,以分裂法产生初始码书(以4×4块图像组成的256码书),以100个图像为训练样本,通过全搜索算法进行码字搜索,对64×64的灰度图像进行压缩。计算还原结果表明,该方法可以得到较高的数字图像压缩比。文章具体研究了训练样本个数与运算时间的关系,计算结果表明,减少训练样本数可以大大缩短编码的运算时间。

Abstract:

A digital image compression encoding method was realized and discussed in this paper. Under the conditions, which the mean square error method is used to determine the distance, the splitting method is used to set the initial codebook (256 codebook build on 4×4 image block), the full search method is employed to ferret codeword, and the sample size is 100 images, the 64×64 grayscales are compacted by LBG compression encoding method. The relation between the number of sample and the run time was studied. The computer result showed that the run time cutting down was to become more remarkable with the decrease amount of sample.