在纹理图象分割的研究领域中,基于人类视觉感知特性的图象分割方法是一个重要的新研究方向.文中给出了一种纹理图象分割的新方法,即将小波变换和方向滤波有效地结合了起来.该方法是根据小波变换的多尺度、多频道滤波的特点,采用小波包提取纹理主频,然后用一种二维最佳正交极可分方向滤波器对纹理有关方向和频率等其他特征进行分割.实验结果表明,这种方法能够获得较好的分割效果.

In the field of texture image segmentation, method based on visual perception become a new effective way to solve this kind of problem. This paper just does such research work on texture image segmentation. In the paper an approach which combines wavelet transform with orientational filtering is proposed. First wavelet package can be used to extract main frequency of the texture since wavelet transform has the character of multi-channel filtering. And then we use a 2D orientational filter to extract the others information about the orientation and frequency, etc. Finally we use a human texture image in the experiment to test the approach of this paper. And we get the processed result of the image. The experiment shows the effectiveness of this method.