纹理图像分割是图像处理的一个基本问题。由于基于高斯混合模型的纹理图像分割方法．大多采用单像素的方法，因此分割精度和效率都较低。为了更好地进行纹理图像分割，在子空间思想的基础上，提出了一个基于图像块的分割算法及其改进算法，即先取图像块的均值、标准差、最大值、最小值以及中间像素的像素值等5个特征作为纹理特征，再利用高斯混合模型进行纹理图像分割，实验结果表明，该新算法的分割精度和分割效率较原分割算法都有较大提高。

In the field of image processing, Image segmentation is the basic problem. Using mixtures of!Gauss, Some people segment image by single pixel value and get poor precision and lower efficiency. For segmenting image well, in this paper, we present a texture image segmentation algorithm by image patches. It is inspired by sub space on some authors. Our experiences show that this algorithm can segment texture image a little, although it cost very much in time. Then we take mean, stand deviation, maximum, minimum and middle pixel value of image patch as features. Our algorithm segments texture image very well. Especially, it improves a lot in time.