Abstract: This paper proposes a new image compression algorithm, based on dynamic side-match finite-state vector quantization and S+P transformation. The algorithm decomposes the original image by the S+P transform, the correlation of the three directions’ wavelet coefficients is used to construct the band-cross vector, and the vector is quantized by the DSMVQ and PNN. Through the method above, the high coding efficiency and reconstructed image quality are both obtained. Results of the experimentation show that this method can be realized easily and compress the wavelet image efficiently under lower coding rate.

Key words: S+P transformation, Dynamic side-match finite-state vector quantization(DSMVQ), Construction of the band-cross vector

摘要： 提出了一种用于图像压缩的动态边匹配有限状态矢量量化(DSMVQ)算法，该算法通过S+P变换方法，对图像进行多级小波变换，利用3个方向上各自小波系数之间的相关性，构造符合图像特征的跨频带矢量，利用DSMVQ进行量化，采样了基于人眼视觉特性的加权均方误差准则，提高了图像的编码效率和重构质量。试验结果表明，该算法实现简单，在较低的编码率下，可达到较好的压缩效果。

关键词: S+P变换, 动态边匹配有限状态矢量量化, 跨频带矢量构造