Multiresolution edge detection algorithm applied to SAR images

Abstract

A multiresolution based algorithm with optimal thresholding method is presented to detect edges in remote sensing images. The images are filtered with high pass and low pass wavelet filters. An overcomplete expansion is constructed. The coefficients are then evaluated for presence of noise. This is specifically useful for SAR images where speckle noise is present and its removal prior to application of our algorithm is not mandatory. The estimate is then used to denoise the image. Example edges from SAR images are given and analyzed. Also the edges are used along with the co-occurrence algorithm for segmentation of the image.