Segmentation of Synthetic Aperture Radar (SAR) Images of Ocean Surface by the Texture Energy Transform Method

A texture energy transform approach has been chosen for the study of texture analysis and classification in digital images. It is based on the idea of detecting the degree of similarity between the spatial variation of image pixels and a set of chosen mask filters. Feature measures which characterize the texture are simpler to compute as compared with other approaches based on auto- correlation functions, digital transform methods, spatial gray tone co- occurrence probabilities, auto-regression models, etc. An algorithm was developed for the segmentation of SAR ocean surface images into regions with and without directional streaks. Excellent results are obtained in which surface areas with relatively rough and calm water are identified.