|Аннотацiя:||Efficiency of Synthetic Aperture Radar (SAR) image despeckling is assessed using model data that take into account basic properties of real-life single-look images, in particular non-Gaussian probability density function of fully-developed speckle and its spatial correlation. Analysis is performed for a wide set of well-known despeckling techniques and for recently proposed locally adaptive filters that combine a Level Set method used to detect small-sized objects and Discrete Cosine transform based denoising. The despeckling performance is evaluated in terms of standard criterion (output MSE) and two other criteria - local MSE in heterogeneous regions and integral MSSIM. The experiments have demonstrated that the locally adaptive filters outperform the well-known ones.
Key words: single-look SAR image; combined despeckling; Level Set method, DCT, efficiency analysis.