基于小波变换和误差竞争学习的矢量量化

Vector Quantization Based on Wavelet-Transform and Distortion Competitive Learning

用神经网络实现图像矢量量化是一种非常有效的方法 ,而小波变换又是近年来迅速发展起来的新算法。文中提出一种改进的误差竞争学习算法 ,分析了图像在小波变换后数据的特点 ,提出了新的矢量构造方法 ,从而最终得到了基于小波变换和误差竞争学习的矢量量化图像压缩新算法 (以下简称VQWDCL) ,无论是在主客观效果上 ,还是在计算复杂度上 ,其性能都优于传统的基于小波变换和LBG算法的矢量量化。

Neural network is a very efficient method for vector quantization, and wavelet transform is a new algorithm developed rapidly in recent years. In this paper a kind of modified distortion competitive learning algorithm was proposed, image data after wavelet transform was analyzed, a new method of vector construction was proposed, and then a new vector quantization algorithm for image compression based on wavelet transform and distortion competitive learning (VQWDCL) was proposed, which is superior to the conventional vector quantization based on wavelet transform and LBG algorithm both on the experiments results and on the computation complication.