一种基于Contourlet的图像低码率SPIHT算法

Contourlet是一种优于小波的不可分离图像表示方法，它除了具有一般小波变换的多尺度、局部性外，还具有方向性、各向异性。为此，本文提出了一种基于Contourlet变换的SPIHT编码方法。首先，对图像Contourlet变换各子带系数的分布情况进行了统计分析；进而设计了一种基于Contourlet变换的空间方向树结构，该结构具有类似图像小波变换域中“零树”的特性；在此基础上实现了一种基于Contourlet变换的渐进式编码算法（CSPIHT），该算法除了具有一般基于小波变换的零树编码算法的特性外，还具有方向性和各向异性的特点，特别对图像的边缘方向信息和纹理信息具有很好的捕捉能力和解码的视觉效果。实验结果表明，所提出的算法CSPIHT在低码率下其重构图像的PSNR要高于SPIHT算法；而对于中等码率，尽管重构图像的PSNR略低于SPIHT，但重构图像纹理和边缘区域的视觉效果要优于SPIHT。

A Low Bit-Rate SPIHT Algorithm Based on the Contourlet Transform

In this paper, a new non-linear image approximation method that decomposes images both radially and angu- larly is proposed. In order to explore the potentiality of this new transform as a tool for image coding, the developed SPIHT coding scheme that is based on using non-linear approximation of images. Through careful statistical analysis on the independent sub-hand coefficients of contourlet transform, the spatial orientation trees based on contourlet trans- form, and this structure has the "Zero-tree" characteristic in wavelet transform domain is proposed. Based on this, the zero-tree encoding (CSPIHT) based on contourlet transform is proposed. This algorithm not only has multi-scale fea- ture, but also has direction and anisotropy, consequently which can enhance the capability of capturing textures, con- tours and fine details in images. In this work, some study and analyze the contourlet transform for bit-rate image cod- ing. Simulation results prove that the proposed scheme(CSPIHT), at the lower decode bit rate, can achieve higher PSNR than traditional SPIHT image compress algorithm; Although the PSNR of decoded image lower than SPIHT at medium bit rate, our algorithm is visually superior to the SPIHT in textures and contours. So this new coding approach is competitive to the wavelet coder in terms of the PSNR-rate curves, and is visually superior to the wavelet coder for the mentioned images.