摘要
将Gabor滤波器和各向异性扩散方程相结合,提出了一种基于活动围道的无监督纹理图像分割算法.采用基于总变分流的扩散函数,各向异性扩散方程可以有效地在保留纹理图像大尺度边界信息的同时,对图像纹理区域进行平滑,获得比原始图像更易分割的简化图像.但是平滑过程中纹理信息的丧失,限制了该方法的通用性和有效性.为了在利用各向异性扩散方法的同时,有效地提取和利用纹理信息,我们利用Gabor滤波器提取一组表征纹理方向性和尺度性的特征图像;同时将原始图像作为表征纹理灰度信息的一个特征通道考虑;再利用矢量形式的各向异性扩散方程,对特征图像进行边界保持的各向异性平滑.将基于区域灰度统计参数估计的活动围道分割方法扩展到矢量空间,来对平滑后的纹理特征量进行分割.实验证明,利用该纹理分割算法可以获得较好的效果.
In this paper, an unsupervised texture image segmentation method based on active contour is presented. Gabor filters and the anisotropic diffusion method are combined to obtain the texture feature for segmentation. A small set of Gabor filters are used to extract scale and orientation properties of the texture , and original image is also included as a feature channel for the intensity information of the texture. If the diffusion function based on total variation is chosen, vector anisotropic diffusion can smooth the details away with the region border reserved in every channel. Finally the geometric MDL active contour image segmentation algorithm is expanded to vector-value data to segment the features vectors. Experiments on various kinds of texture images show that the method is effective.
出处
《中国科学院研究生院学报》
CAS
CSCD
2005年第5期624-630,共7页
Journal of the Graduate School of the Chinese Academy of Sciences
