ON SEA-ICE TEXTURE CHARACTERIZATION FROM SAR IMAGES

Automated sea-ice classification from radar images has been identified as one of the prime goals of future sea-ice monitoring programs. The performance of any image-classification scheme depends on the image parameters selected to characterize the existing classes. Therefore a search for parameters that promise strong interpretation capabilities for radar images is a priority at present. This paper reports briefly on a study to evaluate the utility of selected texture parameters, derived directly from the gray level or indirectly via the gray-level co-occurrencematrix (GLCM), for sea-ice classification from SAR images. Two images (one L-band and the other X-band) of winter sea ice were used. The paper shows that texture can be used to identify certain sea-ice classes, particularly multiyear and new ice. Of the six texture parameters investigated, the gray-level variance and the GLCM Entropy and Uniformity seem to be the most significant. © 1990 IEEE