Texture segmentation based on MRMRF modeling

摘要

Texture segmentation remains a fundamental issue in low-level image analysis, pattern recognition and computer vision. Texture segmentation problem can be solved in two directions: model fitting and non-parametric classification. In this paper, we propose to use multi resolution MRF (MRMRF) modeling in texture segmentation. A novel MRMRF parameter estimation method based on MCMC approach is presented. The experimental result shows that the method is suitable to segment textured images.