一种小波树分类和合成编码结合的图像压缩方法

A Method of Image Compression Using Wavelet Tree Classification and Hybrid Coding

针对传统小波变换过程复杂的缺点和SPIHT算法编码过程重复运算、存储量大的问题,提出了一种小波树分类和合成编码结合的图像压缩方法.该方法首先对纹理丰富的图像进行3级小波变换,再对高频子带通过小波树分类器分为低频树和高频树.最后对最低频子带进行差值脉冲编码调制(DPCM),对低频树和高频树分别进行SPIHT和多阶段矢量量化(MVQ.)仿真结果表明,该方法在峰值信噪比和编解码时间均优于SPIHT算法.在0.125 b/s下,Boat图像峰值信噪比(PSNR)比SPIHT算法提高了0.8 dB.

In view of the problems of complicated convolution process of wavelet transform, repeated calculations and a huge capacity of memory in SPIHT algorithm, a method of using a combination of wavelet tree classification and hybrid coding for image compression was presented. First, wavelet transform is applied to rich image for a full de-composition. Next, a wavelet-tree classifier can efficiently divide the high frequency subbands into low frequency tree and high frequency tree. Finally, the DPCM, the SPIHT and the MVQ are used to code in the lowest frequency subband, for low frequency tree and high frequency tree respectively. Simulation results show that this method can be better than SPIHT conceming the PSNR and coding/decoding time. The PSNR of Boat image increases by 0.8dB using this method at 0.125b/s.