Locally adaptive edge preserving filter for radar image denoising

This paper presents a locally adaptive filtering technique for denoising radar images corrupted by intensive speckle. The proposed filter is based on a combination of discrete cosine transform (DCT) and tetrolet filter applied after variance stabilizing transformation. Through optimization of filter parameters, the edge preservation capability of the tetrolet based filter for the heterogeneous image region alongside with the effectiveness of the DCT based filter for the homogenous image region is shown to improve the visual quality of denoised images according to several visual quality metrics.