Comparative Analysis of Noise Reduction Techniques for Brain MRI Images

Comparative analysis of noise reduction techniques for brain MRI images delves into the evaluation of various noise types, including salt and pepper, gaussian, and speckle noise. This study considers three prominent filters: the median filter, non-local means (NLM) algorithm, and gaussian filter, along with performance metrics such as Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), Mean Squared Error (MSE), and Entropy. When confronted with salt and pepper noise, the Non-Local Means (NLM) filter excels, reducing noise with minimal information loss. However, Visual inspection favors the median filter for effective noise reduction. For speckle noise, a comprehensive assessment combining quantitative analysis and visual inspection favors the NLM algorithm, showcasing its noise reduction capabilities without compromising image information. In the case of gaussian noise, quantitative analysis underscores the NLM filter's ability to achieve high PSNR values, emphasizing its proficiency in noise reduction, though with noticeable smoothening. For Gaussian filter, while blurring the image, retains information effectively. In summary, the NLM filter is robust for speckle and gaussian noise, excelling quantitatively and visually, while the median filter is preferred for salt and pepper noise when preserving image details is crucial. This analysis provides insights for informed decision-making in enhancing brain MRI image quality.