一种基于EM算法的图像分割改进技术

A Modified Method of Image Segmentation Based on EM Algorithm

图像分割是计算机视觉的基础，该文结合EM算法和PCA降维技术，给出了一种有效快速的进行图象分割的方法。该方法利用高斯混合模型对原始图像进行建模，通过EM算法将分割问题转化为参数最大似然估计的问题，同时采用PCA降维技术和随机采样来降低计算量。通过人工合成图象及真实图象的实际测试结果，验证了该算法的有效性和快速性。

Image Segmentation is an elementary problem in computer vision. In this paper, an efficiently method of image segmentation was proposed by using EM algorithm and PCA dimensionality reduction algorithm. Firstly a gaussian mixture model（GMM） is employed to modeling the image, and then get GMM＇s parameters by using EM ,In order to reduce complexity, Some methods for reduce feature dimensionality and data points is used. Experiments results on natural images and composite images show that the method is effective