Combined Speckle Removal Based on the Wiener Filter and Wavelet Transform using Texture Analysis

The existing despeckling methods typically encounter difficulties in achieving an optimal equilibrium between the suppression of noise and the preservation of critical image features, particularly in scenarios involving images that encompass both homogeneous (smooth) areas and heterogeneous (textured or edge-dense) regions. This paper proposes a hybrid approach combining wavelet filtering and Wiener filtering, leveraging their complementary strengths. The results of computer modeling have shown that the wavelet filters are effective in homogeneous regions by exploiting multi-scale sparsity, while the Wiener filter adaptively preserves textures in heterogeneous areas. By segmenting the image based on regional homogeneity, we apply the corresponding filter, achieving superior noise reduction in the sense both objective and subjective criteria.