Combining ICA with LSSVM for Speckle Reduction of SAR Image

Speckles are inherent defects due to coherent imaging principles in synthetic aperture radar (SAR) systems. An improved polarimetric SAR image filtering method based on ICA (independent component analysis) -LSSVM (least squares support vector machine) was developed to get better results for terrain classification, target detection and other applications. The first step of this method is to apply ICA to LSSVM for feature extraction. By using ICA, the original higher dimensional inputs will be transformed into other lower dimensional features, the regularized parameter and feature subsets will be geted,These new features are then used as the inputs of LSSVM to process a regularized procedure which defines a cost function to estimate true SAR image.Experiment results using polarimetric SAR data of ERS-1 satellite show that this algorithm attains good despeckling effect and preserves more valuable image details than traditional algorithms.