Texture segmentation using filters with optimized energy separation

The design of filters for texture feature extraction is addressed. Based on a new feature extraction model, optimization approaches utilizing various feature (energy) separation criteria are developed. Both two- and multiple-texture problems are addressed. The approaches are assessed by supervised segmentation experiments. The experiments also include results from alternative filter optimization approaches.