基于综合特征的纹理分割方法

Texture Segmentation Method Based on Integrated Features

针对灰度共生矩阵只能在单一尺度对纹理进行分析的不足,结合离散框架小波变换产生的尺度共生矩阵与梯度变换图像的灰度共生矩阵,提出了一种具有多尺度分析特性的综合纹理特征提取算法,并利用该特征对纹理图像进行分割。仿真实验结果表明:与基于单一尺度特征的纹理分割方法相比,本文提出的算法能够提高纹理边界定位准确性,减少区域内像素错分,取得了较好的分割效果。

To overcome the shortage of the texture analysis based on gray-level co-occurrence matrix（GLCM） in single-scale, an integrated features extraction algorithm is proposed and demonstrated with application to texture image segmentation, which combines the GLCM of gradient image and the scale co-occurrence matrix of discrete wavelet frame transform（DWFY） and has multi-scale analysis characteristics. Experimental results indicate that the method can improve the precision of edge location, reduce the false classification rate, and get better segmentation than the one based on the single-scale features.