Performance Analysis of Non-Local Means Denoising on Medical Images and the Impact of Filter Parameter Variation

In medical imaging, diagnostic accuracy depends on clear images, which are frequently damaged by noise. To address this researchers have developed a number of noise reduction techniques, each with unique benefits and limitations. The effectiveness of NLM (Non-Local Means) denoising in medical images at various noise levels is comprehensively examined in this work. The goal is to understand how key parameters patch size, window size, and smoothing parameter (h) interact under different noise conditions. MSE (Mean Square Error), PSNR (Peak Signal-to-Noise Ratio), and SSIM (Structural Similarity Index) metrics are applied in performance evaluation to provide quantitative insights into the efficiency of denoising and structural preservation. Experiments are conducted using a diverse dataset that includes MRI (Magnetic Resonance Images) and CT (Computed Tomography). Throughout the study, consistent patch sizes of 3 and 5, along with a window size of 11 and 15, are used to explore the role of the smoothing parameter for different noise levels and assess denoising efficacy at various noise intensities. This study contributes to understanding the potential of NLM denoising in improving image quality under varying noise situations, emphasizing the significance of deliberate filter parameter selection over random choices for improved denoising in both objective and subjective approaches.