Mixture Models and the Segmentation of Multimodal Textures

A problem of using mixture-of-Gaussian models for unsupervised texturesegmentationisthat "multimodal"textures(such ascan often be encountered in natural images) cannot be well represented by a single Gaussiancluster. We propose a divide-andconquer method that groups together Gaussian clusters (estimated via Expectation Maximization) into homogeneous texture classes. This method allows to succesfully segment even rather complex textures, as demonstrated by experimental tests on natural images.

• Type: Conference
• Year: 2000
• Where: CVPR
• Authors: Roberto Manduchi
• Keywords: Complex Textures, Computer Vision, CVPR 2000, Expectation Maximization, Gaussian Clusters, Homogeneous Texture Classes, Natural Images