基于小波系数均衡方向变换的图像压缩

Uniform directional transform based on wavelet coefficients for image compression

本文提出了一种新的基于小波系数的均衡方向变换方法,首先用小波变换对图像做多尺度分解,再用改进后的均衡方向滤波器组(uniform directional filter banks,uDFB)对小波变换后的高频子带作方向变换。该变换是对基于小波的contourlet变换的一种改进,它充分利用了小波的非线性逼近特性以及均衡方向滤波器良好的多分辨率和多方向特性,进一步增强了图像的方向特征,同时能有效抑制瑕玷的引入。仿真结果表明,将该方法用于图像压缩可以比基于小波的压缩算法更好地恢复出原图像的轮廓和纹理,并获得较高的峰值信噪比。

This paper proposes a new uniform directional transform method based on wavelet coefficients. First, the wavelet is used for image multiscale decomposition, and then the directional transform is added to the high frequency wavelet subband followed by the improved uniform directional filter banks. This transform is an improvement of wavelet coefficient based on contourlet transform, not only it takes advantage of the wavelet transform that has efficient nordinear approximation property, but also the uniform directional filter banks that has multiresolution and multidirectiona property. Thus, it can enhance the important feature of directionality to image and suppress the artifacts. The experimental results show that when applied to image compression, this method can restore the contours and textures better than the algorithm based on wavelet transform, and it can also get higher peak signal to noise ratio （PSNR）.