无抽样方向滤波器组的设计及其在医学图像增强中的应用

Design of decimation-free directional filter bank and its application in medical image enhancement

本文提出一种无抽样的方向滤波器组设计方法,并应用于医学图像增强。该方法将多级滤波器组整合为单级滤波器组,并将各频域滤波器均转化为w×w的空间模板,分别与图像卷积,获得各方向的子带图像,且每幅子带图像的分辨率均与原图像相同,便于利用各方向子带图像的统计特性,实现图像增强。实验结果表明:本方法因避免了抽样与插值过程,操作简便。采用图像方向特征,降低由于设备问题造成的图像模糊的影响,与其他常见的图像增强方法相比,其图像增强效果明显。

The paper presents a new directional analysis tool namely decimation-free directional filter banks and their operators, as well as applies the operators to medical image enhancement. It integrates multilevel filter banks into single level filter bank and transforms all directional filters in frequency domain into spatial operators with the size of w x w. Each subband image with the same size as the input image can be obtained by convolving input image with each operator. We can get enhanced image according to the statistical character of all subband images. Artifacts due to aliasing and folding during downsampling are avoided in decimation-free directional filter bank due to absence of decimators. No interpolation prior to image synthesis is required, so artifacts produced due to interpolation can be avoided. Results proved that using directional information can reduce image blurring. The method outperforms other classical methods in terms of visual quality, especially in texture images.