A Hybrid Approach for CT-MR Brain Image Denoising Using PSO-Optimised Non-Local Means and Wiener Filtering

Abstract

Noise in CT-MR brain images poses a critical challenge, significantly impacting diagnostic accuracy as well as clinical decision-making. Current medical image-denoising techniques struggle to effectively remove noise while preserving crucial image features. The hybrid technique's potential that combines complementary denoising algorithms is highlighted by the limitations of these approaches. This paper develops & evaluates a denoising method that combines the strengths of Particle Swarm optimised, Non-Local Means (NLM) & Wiener filtering. The proposed approach effectively denoises by leveraging non-local self-similarity in brain images. Particle Swarm Optimisation algorithm is employed to fine-tune the smoothing parameter of NLM denoising, ensuring optimal performance, while the Wiener filter helps to address the trade-off between noise reduction as well as edge preservation. The proposed denoising technique outperforms the traditional methods including median, gaussian, NLM, and wiener filter in terms of peak signal-to-noise ratio (PSNR), image quality index (IQI), mean square error (MSE), and structural similarity index (SSIM).