A PROPOSAL FOR EDGE DETECTION IN SAR IMAGES

MIGUEL A. RE; GUILLERMO G. AGUIRRE VARELA; SERGIO MASUELLI

Congreso; VII MACI 2019 - VII Congreso de Matemática Aplicada, Computacional e Industrial; 2019

It is presented a method for detecting borders in continuous range sequences. It is proposed as an edge detector in SAR images. The method is based on Jensen Shannon divergence (JSD), an entropic distance measure. It is defined a sliding window over the sequence and the probability densities are approximated by a kernel density approximation. In this way, unlike usual methods reported in the literature, we make no use of any particular model for the speckle. There are shown the results obtained in the detection of water ice edge in an antartic region and the values of JSD when moving along the rows in the image.