MULTIFRACTAL DIMENSION AND ITS GEOMETRICAL TERRAIN PROPERTIES FOR CLASSIFICATION OF MULTI-BAND MULTI-POLARIZED SAR IMAGE

Multifractal dimensions Dq for real q are a more general parameter than the fractal dimension in describing geometrical properties. It has been shown that the four multifractal dimensions D-1, D0, D1 and D2 are able to extract different surface information of SAR images. In this paper, we investigate the dimension properties of multifractal dimensions. For land use classification where the textural information on the surface is important, it is necessary to look into the properties of multifractal dimensions with the geometrical properties of terrain. In order to extract the surface information from SAR images, the optimum number of multifractal dimensions to be used in the classification process is considered. To address the suitability of these parameters, these parameters are applied on a multi-band SAR image with regions of different textural information and the results are studied. The abilities of multifractal dimensions in extracting information for different land use classes are considered. In general, although multifractal dimensions provide additional information about the land use classes, there is no clear relation among the land use classes, image polarization and multifractal dimensions.