图像编码的多分辨率分类矢量量化算法

An Algorithm of Multiresolutional Classified Vector Quantization for Image Coding

本文结合小波多级变换后不同方向、不同尺度的系数相关性，提出了一种多分辨率的分类矢量量化方案，对由不同分辨率水平、不同方向的小波系数组成的矢量块进行分类量化编码，并设计了相应的性能良好的分类器，给出了不同条件下的实验结果和数据，从而证明了该编码方案不但提高了图像的质量，同时计算复杂度也有一定的降低．

This paper suggests an algorithm of multiresolutional Classified Vector Quantization for Image Coding,in according to the correlation among the various scales and different orientation bands after wavelet transformation.This algorithm rearranges the coefficients from various resolution level and orientation and gets a vector of 64 dimensions which then is classified and quantized.Meantime,we design an efficient classifier,and give the experiment results and data under different conditions.Compaered with other methods,it proved that the algorithm not only improves the image quality,but also decreases the computation complexity.