Merging spectral and textural information for classifying remote sensing images

Machine interpretation using pattern recognition technique for remote sensing images not only liberates plenty of human resources, but improves results on the efficiency in certain aspects. In this paper, land/water classifiers that combined the texture measures with spectral analysis for remote sensing images have been built. And specific recognition has been designed in order to take advantages of both analyses. The use of red/infrared spectral analysis refines the boundary of land/water; meanwhile the merging of co-occurrence matrix texture analysis and spectral information has improved the accuracy of the two-class labeling.