PERFORMANCE EVALUATION OF TEXTURE MEASURES FOR GROUND COVER IDENTIFICATION IN SATELLITE IMAGES BY MEANS OF A NEURAL-NETWORK CLASSIFIER

The performance of several feature extraction methods for classifying ground covers in satellite images is compared. Ground covers are viewe d as texture of the image, Texture measures considered are: cooccurren ce matrices, gray-level differences, texture-tone analysis, features d erived from the Fourier spectrum, and Gabor filters, Some Fourier feat ures and some Gabor filters were found to be good choices, in particul ar when a single frequency band was used for classification, A Themati c Mapper(TM) satellite image showing a variety of vegetations in centr al Colorado was used for the comparison. A related goal was to investi gate the feasibility of extracting the main ground covers from an imag e, These ground covers may then form an index into a database, This wo uld allow the retrieval of a set of images which are similar in conten ts, The results obtained in the indexing experiments are encouraging.