Textural information in SAR images

Image variance for a given land use category is presently related to individual variances associated with image speckle and target texture by means of a multiplicative model; in it, speckle is treated as a random process governed by signal fading, that is independent of the textural variations associated with scattering property spatial variations in visually uniform 'distributed' targets. Seasat SAR imagery of Oklahoma was used to evaluate the textural autocorrelation function of water, forest, pasture, urban, and cultivated land use categories. It is found that the image contrast and inverse moment second-order statistics furnished a classification accuracy of 88 percent, with only modest spatial resolution degradation. A second study based on SIR-A imagery of five forested regions shows that the use of textural data can improve classification accuracy from 75 to 93 percent. (Author)