INVESTIGATION OF DIFFERENT METHODS OF THRESHOLD SELECTION FOR DCT-SSA FILTER

Abstract

Synthetic aperture radars (SARs) are used to acquire images that are employed for a variety of practical applications, from analyzing soil and vegetation cover to searching for large accumulations of illegally dumped debris. The problem of processing such images is that the obtained images are spoiled by speckle - a noise-like effect with multiplicative nature. Also, speckle has a non-Gaussian distribution and is spatially correlated. To address this problem, many filters have been developed that take into account the non-Gaussianity of the distribution and also provide acceptable speckle suppression. Discrete Cosine Transform (DCT) based filters or BM3D filters are examples of such filters. A common problem with most filters remains that they distort low-size details, edges of large-size objects and textures, resulting in an undesirable blurring effect. This paper tests the assumption that by changing the type of threshold and the method of its calculation for image blocks, it is possible to achieve improved preservation of point-like objects and other informative features. The results of a study using determination of the local threshold with calculation of mean, median and minimum interquantile distance for hard and combined thresholds are presented. The DCT-SSA filter is chosen as the basis, since DCT has good decorrelation properties, and the filter itself has repeatedly demonstrated quite high speckle suppression efficiency. PSNR, PSNR-HVS-M and HaarPSI are used as quality metrics, where the latter one is first applied to evaluate the efficiency of SAR image filtering. The data obtained in this study indicates that the proposed modifications result in better preservation of bright points in the image and can be used to improve the image quality of the DCT filter output.