Adaptive Median-Based Clustering Approach to Restore MRI Images from High-Density Salt-and-Pepper Noise

In real-time medical imaging, it is crucial to extract detailed feature sets from high-density noise to recognize patterns that are indicative of disease symptoms. The noise present in MRI images introduces various errors, resulting in a grainy texture. In medical image processing, salt-and-pepper noise randomly changes the image pixels into isolated 0(s) or 255(s). A cluster-based analysis is proposed to achieve an accurate approximation of noise-free pixels in MRI images. Consequently, this study introduces the concept of high density to accurately estimate pixels alongside values of zero or 255. This is because MRI images frequently contain image pixels of 0 and 255. In this algorithm, the filter initially functions as a Standard Median Filter. However, when the densities of 0 and 255, along with salt-and-pepper noise, increased to approximately 40\% of the total pixels in the mask, the clustering process was initiated by creating a kernel size of (11×11). The pixels were divided into two groups, assuming k = 2. To validate the effectiveness of the proposed filter, it was compared with existing state-of-the-art filters, and its efficacy was confirmed using MRI and grayscale images. The simulation results of the peak signal-to-noise ratio and structural similarity index demonstrate that the proposed algorithm outperforms other filters in rendering noise-free medical images.