Fusion of Evidences in Intensity Channels for Edge Detection in PolSAR Images

Polarimetric synthetic aperture radar (PolSAR) sensors have reached an essential position in remote sensing. The images they provide have speckle noise, making their processing and analysis challenging tasks. We discuss an edge detection method based on the fusion of evidences obtained in the intensity channels hh, hv, and vv of PolSAR multilook images. The method consists of detecting transition points in the thinnest possible range of data that covers two regions using maximum likelihood under the Wishart distribution. The fusion methods used are: simple average, multiresolution discrete wavelet transform (MR-DWT), principal component analysis (PCA), receiver operating characteristic (ROC) statistics, multiresolution stationary wavelet transform (MR-SWT), and a multiresolution method based on singular value decomposition (MR-SVD). A quantitative analysis suggests that PCA and MR-SVD provide the best results.