基于GLCM和EM算法的纹理图像分割

Texture Segmentation based on Gray-Level Co-occurrence Matrix Features and EM Algorithm

基于纹理图像的特征,提出了基于灰度共生矩阵(GLCM)和快速极大似然估计(EM)算法相结合的纹理图像分割新算法,为了获得较好的纹理图像分割结果该算法采用灰度共生矩阵的三个常用特征并在四个方向上求平均,从而克服了方向的影响。采用欧式距离度量函数求得两特征向量的距离。通过用改进EM算法对距离矩阵进行聚类,得到纹理图像的初始分割结果,最后用形态学的方法实现对纹理图像边界的精确定位。

This paper proposes a textures segmentation algorithm based on gray-Level co-occurrence matrix features and EM.Three texture features is computed,thus to gain better result.The average measurements of all four angles are usually used in calculating GLCM for reducing the influence of angle.The distance between two texture feature vectors are measure.To derive the initial segmentation result,distance matrix is used as the feature vector and the improved expectation maximization image segmentation is applied to achieving effective segmentation.Finally,by using the method of morphology,the accurate localization of region boundaries is realized.