An Interpretable SAR Image Filtering Algorithm

Effective noise suppression is crucial for the subsequent interpretation tasks of synthetic aperture radar (SAR) imagery. Traditional SAR image processing techniques often overlook the coherent nature of noise, leading to a loss of vital detail during filtering. With advancements in deep learning (DL), significant strides have been made in image processing. However, existing DL methods do not fully leverage the imaging mechanisms of SAR, resulting in a lack of specificity and interpretability in the filtering process. To balance noise reduction with detailed preservation and to address the “black box” issue in filtering, we propose an interpretable filtering method that employs a correlation-based upward search for density peaks. Initially, we develop a MeanShift-Markov random field (MS-MRF) filter that integrates MeanShift with MRF in the joint spatial–spectral domain, ensuring both correlation and detailed preservation; the derivation of the MS-MRF filter is rigorously grounded in mathematical theory. Subsequently, we integrate MS-MRF with convolutional operations in DL to create a novel convolutional filter, interpretable MS-MRF convolution (IMMC), which enhances the model’s interpretability, noise reduction capabilities, and detailed retention. Extensive experiments demonstrate that our method outperforms state-of-the-art (SOTA) SAR denoising techniques, achieving an average structural similarity index (SSIM) of over 85.00% and an average peak signal-to-noise ratio (PSNR) exceeding 35.00 dB across synthetic datasets with varying noise levels, showing significant improvements in noise suppression, detailed preservation, and interpretability.