Texture segmentation method by using two-dimensional AR model and Kullback information

摘要

This paper deals with a segmentation method of texture image with randomness. For the texture segmentation, the statistical characteristics of texture are extracted by a texture measure, and it is necessary to discriminate textures by using the measure. In this paper new discriminant functions based on two-dimensional autoregressive models fitted to the texture image being considered and Kullback information which can measure the distance between two probability distributions are proposed. Furthermore, the texture segmentation method using this function is proposed. By numerical examples the validity of the method is verified.