Segmentation of satellite images in optoelectronic system

The problem of segmenting the satellite images into homogeneous texture regions that correspond to the different classes of terrestrial surface is considered. It is shown that this problem may be successfully solved by using the method of spectral synthetic discriminant functions recently proposed by the authors for classification of random image fields and realized by means of a rather simple optoelectronic technique. The experimental results of segmenting the true satellite images are given.