MRI Image Denoising Influenced By Iterative Hint Pixel Searching Through Multi-Order Neighbours And WSDHM Based Dimensionality Reduction

Abstract

An essential tool in medical diagnostics is Magnetic Resonance Imaging (MRI), but its effectiveness is often compromised by noise, particularly impulse noise (or salt-and-pepper noise), which can obscure vital image details. This research proposes a denoising method entitled ‘MRI image denoising using Iterative Hint Pixel Searching through Multi-Order Neighbors and WSDHM-based Dimensionality Reduction (IHSMW)’, which uniquely combines a new hint pixel searching algorithm and a new hint-data dimensionality reduction algorithm.  A novel algorithm for hint pixels searching is designed which is named ‘Four Corners based Multi-order supported Iterative Hint Pixel searching algorithm (FCMIHP)’. A new data dimensionality reduction technique known as Weighted Shortest Distance Hierarchical Mapping (WSDHM) that significantly improves the denoising process is also designed. The proposed IHSMW filter excels in preserving essential structural details in grayscale brain MRI images, thus enhancing the diagnostic potential, especially in brain cancer detection and segmentation. The proposed method is validated against traditional denoising techniques and demonstrates superior performance in preservation of crucial image components while noise is significantly reduced. The experimental findings reveal that the proposed IHSMW method using the PRNV-DB database gives a better PSNR value of 31.55% and the lowest MSE value of 45.507%, which means it gives a better result than existing methods.

Keywords:  , ,.