关键词: 高斯图模型，变量选择，L1惩罚正则化，彩色纹理分类

Abstract: Texture is one of the important visual features in image analysis. For convenience, color texture images are often converted to gray images. It is a pity that color information is ignored. In order to keep texture and color informa- tion,principle component analysis (PCA) was utilized to reduce the dimension of color textures. Gaussian graphical models (GGM) have good prospect due to themselves advantages, and are applied to construct texture model. The structure of GGM is explored by the connection between the local Markov property and conditional regression of Gauss- ian random variables. Thus, the model selection can be converted to select variables in GGM. The development of tech- nic}ue of penalty regularization provides many methods for variable selection and parameter estimation. And, the methods of penalty regularization conduct neighborhood selection and parameter estimation simultaneously. Then, the texture feature is extracted and applied in color texture classification. The experiments show the good results. Therefore, the texture models based connection of GGM and PCA have an attractive prospect.

Key words: Gaussian graphical models, Variables sclection,I:-penalty regularisation, Color texture classification