纯二维全相位滤波器组及其在图像压缩中的应用

True 2-D All Phase Filter Bank and its Application for Image Compression

基于全相位理论,设计了3种纯二维全相位滤波器组(ADFB)。这些滤波器组可以直接对图像进行纯二维分解和重构,若忽略滤波器的有限精度问题,系统是全重建的。为了减小运算量,纯二维ADFB采用提升格式实现。仿真试验表明,纯二维ADFB具有良好的数据压缩特性。在压缩率相同的情况下,尤其是IDCT-AFB7.7比Daubechies9/7小波重建图像的峰值信噪比(PSNR)最多小0.7dB,这与采用的适合于可分离二维小波变换的四叉树SPIHT编码方案有关,若采用二叉树SPIHT编码,会使压缩效果得到改善。

Based on all phase theory,this paper designed three kinds of true 2-D all phase filter bank（APFB） ,which can be used to decompose and recompose image data in true 2-D directly. If ignoring finite precision of filters,the true 2-D APFBs have perfect reconstruction property. To reduce computation, they are implemented in lifting scheme. Simulation has shown that true 2-D APFBs have nicer data compression property. With the same compression rate, the PSNR of reconstruction image of IDCT-AFB7. 7 is at most less 0. 7 dB than which of Daubeehies9/7 wavelet. This is because quadtree SPIHT wavelet coding used in this paper is fit for two dimension separable wavelets. Compression effects will be improved if binary tree SPIHT wavelet coding is adopted.