Hybrid Non-Local Means and Anisotropic Diffusion Filtering for Denoising of CT Brain Tumor Images

Abstract

Medical image denoising is an important progressive step that improves the accuracy of the calculated tomography (CT) brain tumor diagnosis. Existing filtration techniques, such as median, Wiener, and Gaussian filters, struggle with noise removal and balanced protection, compromising clinical quality. This research addresses these challenges with a better hybrid denoising framework, which is adapted to species, salt-and-pepper, poison, and Gaussian noise-affected CT-brain tumors, with a combination of non-local means and anisotropic diffusion filtration. The proposed method makes traditional approaches much better, improving up to 18%in a top signal-to-noise ratio (PSNR), SSIM surplus (SSIM) of 15%, and a reduction in Root Mean Square error (RMSE) up to 22%. Verification against the eight-state Art methods shows better conservation of tumor boundaries and expansion loyalty. These results confirm the framework's ability to detect reliable initial tumors and later clinical decisions, reduce the gaps identified in noise suppression, and improve cute quality retention.