Experiments in Segmenting Natural Textures
This chapter presents results of the supervised segmentation of different natural textured images and artificial collages of natural textures. The conditional Gibbs model of region maps in Eq. (2.26) suggests that to segment a piecewise-homogeneous grayscale texture we have to simply simulate the region maps relating to this texture. Such a segmentation is also based on the CSA that adapts the model potentials as to approximate the GL/RLH and GLC/RLCHs or GLD/RLCHs for a given training sample by the similar histograms for the image to be segmented and its simulated region map.
KeywordsTraining Sample Digital Elevation Model Synthetic Aperture Radar Image Texture Region Test Collage
Unable to display preview. Download preview PDF.