Optimal Condition Evaluation for WNN Based Speckle Filtering Algorithms in SAR Image Denoising

This paper investigates the application of the weighted nuclear norm minimization method for despeckling synthetic aperture radar images, for various input configurations and acquisition scenarios. The approach, originally designed for additive white Gaussian noise, was adapted to handle SAR-specific multiplicative noise using a homomorphic transformation. Speckle filtering was performed in the logarithmic domain, followed by exponential reconstruction to extract the image. Various similarity measures were used to evaluate the performance. Results reveal that despeckling log-transformed images showed superior performance compared to other evaluated methods, achieving better visual quality and structural preservation. While different similarity measures applied to log-transformed images provided comparable results, the WNNM method consistently demonstrated its ability to enhance image quality and objective metrics, and its effectiveness strongly dependent on identifying similar segments in noisy images. The result of the evaluation are presented with extensive simulations results on various test images and SAR data.