Study of Synthetic Aperture Radar Data Compression and Encoding. Part 3. Performance Evaluation of Speckle Suppression and Data Compression Algorithms

摘要

Rate distortion bounds for SAR images are compared with rate versus distortion relations obtained with speckle suppression and data compression algorithms. A method for optimally processing multispectral SAR-images is given. It uses the spectral correlation between the mean return power corresponding to each spectral channel. Real SAR-data is processed with the algorithms and subjected to information extraction experiments. Synthetic SAR-images can not efficiently be processed by the speckle suppression algorithm for one and four looks, if the goal is to obtain a least squares estimate of the reference image. For Seasat imagery (4 looks) data reduction with a compression ratio of 8, without speckle suppression, gives very acceptable results, with almost no impact on image segmentation for land scenes and on Fourier analysis for ocean scenes. The extraction of dominant ocean wave length and direction is not influenced by data compression and speckle suppression applied to Seasat data, even when the compression ratio is 20, and the appearance of Seasat imagery improves if speckle suppression is applied.