基于空间组合矢量分类的小波图像网格编码矢量量化

本文基于零树编码、矢量分类和网格编码量化的思想,提出了对小波图像采用空间矢量组合和分类后进行网格编码矢量量化的新方法.该方法充分利用了各高频子带系数频率相关性和空间约束性,依据组合矢量能量和零树矢量综合判定进行分类,整幅图像只需单一量化码书,分类信息占用比特数少.对重要类矢量实行加权网格编码矢量量化,利用卷积编码扩展信号空间以增大量化信号间的欧氏距离,用维特比算法搜索最优量化序列,比使用矢量量化提高了0.6db左右.该方法编码计算复杂度适中,解码简单,可达到很好的压缩效果.

Trellis Coded Vector Quantiztion of Wavelet Image Using Vector Space-Combination and Classification

Based on zerotree coding,vector classification and trellis coded quantizaion, A new method which performs space-combination vector generation and classification to wavelet image then uses TCVQ is presented. The method utilizes the spase and frequency correlation of the subbands fully, and clsssifies the combined vector through joint judgement using the vector energy and zerotree-vector the whole image only need one quantization codebook, the classified information need a little bites to express. The important vectors are quantified by weighted TCVQ,which uses convolutional coding to expand signal space and increase Euclid distance between signals, uses Viterbi algorithm to find a optimized survived quantized sequence, it has an advantage of 0.6db or so over VQ. The method has modest encoding complexity with simple decoding and it can achieve much better compression effect.