Color Textured Image Segmentation Based on Spatial Dependence Using 3D Co-occurrence Matrices and Markov Random Fields

Image segmentation is a primary step in many computer vision tasks. Although many segmentation methods based on either color or texture have been proposed in the last decades, there have been only few approaches combining both these features. This work presents a new image segmentation method using color texture features extracted from 3D co-occurrence matrices combined with spatial dependence, this modeled by a Markov random field. The 3D co-occurrence matrices provide features which summarize statistical interaction both between pixels and different color bands, which is not usually accomplished by other segmentation methods. After a preliminary segmentation of the image into homogeneous regions, the ICM method is applied only to pixels located in the boundaries between regions, providing a fine segmentation with a reduced computational cost, since a small portion of the image is considered in the last stage. A set of synthetic and natural color images is used to show the results by applying the proposed method.