Genetic algorithm optimization of nonlocal means filter for gaussian noise reduction

Noise reduction is a crucial pre-processing step in imaging. In this study, the nonlocal means filter, known for its edge-preserving properties, was employed to eliminate Gaussian noise. The effectiveness of the nonlocal means filter heavily depends on the selection of its parameters. As an innovative approach, a genetic algorithm was utilized to optimize these parameters. Gaussian noise with several variances was introduced to the images, and genetic algorithm was applied to determine the optimal filter parameters, using peak signal-to-noise ratio (PSNR) as the fitness criterion. Following the optimization, the performance of the nonlocal means filter with the proposed parameters was evaluated and compared with other nonlocal means filter configurations using quality metrics such as mean squared error (MSE), PSNR, and structural similarity index metric (SSIM). The results demonstrated that the nonlocal means filter optimized with genetic algorithm outperformed previously used parameters, achieving superior edge preservation and noise reduction across various levels of Gaussian noise. The obtained results suggest that the efficacy of nonlocal means filter for denoising is highly reliant on the careful selection of optimal parameters.