A LINE-PRESERVING POST-PROCESSING TECHNIQUE FOR MAXIMUM-LIKELIHOOD CLASSIFICATION OF SAR IMAGES

Speckle in Synthetic Aperture Radar (SAR) images causes many pixels to be misclassified when a maximum likelihood classifier (MLC) is applie d. We present a post-processing procedure for removing these misclassi fied pixels. The procedure, requiring only a few comparisons for each pixel, first detects misclassified pixels, then reclassifies them to t he classes of their neighbouring pixels. Experimental results show tha t this procedure efficiently reduces misclassified pixels while preser ving lines, boundaries and corners in thematic maps.