A Method of Finding Optimal Parameters of Speckle Noise Reduction Filters

The reduction of multiplicative speckle noise in synthetic aperture radar (SAR) images is an important problem. Many speckle noise reduction filters have been proposed. Most of them have several parameters that control their operation. Finding the optimal values of these parameters is often a non-trivial task. A method of automating the search for optimal parameters is proposed. The method uses two variants of a specially designed test image, original noise free image and the same image but with speckle noise added. Then the Structural Similarity Index (SSIM) metric is used for finding the parameters that make the filtered image as close to the original noise free image as possible. The application of the method is illustrated using the Frost filter applied to various images, but the method can be used for any filter type.