Performance Analysis of IIC techniques for Brain MR-images
Pooja Thakur, Neelam Syamala, Yepuganti Karuna, Saritha Saladi · 2023
This paper aims to solve the issue of intensity inhomogeneity in brain MR-images. Intensity inhomogeneity causes the development of false pixel intensities within an MRimage making its diagnosis difficult. Such images could even lead to the development of incorrect diagnosis. The issue of intensity inhomogeneity also causes a major issue in image segmentation. Although many standard image denoising methods exist and are used for such correction purposes, these are not all as focused and specific to this issue. In this paper, an attempt to develop a comparative study of a few commonly used standard methods and a newly suggested deep learning solution for this problem is made. All methods given in this paper are compared in the same platform using the same performance metrics to get a fair idea of which methods yield the best results. In this paper, python programming language is used to compare the methods. The methods included in this study are InhomoNET, BM3D, WNNM, NLM Algorithm, FFT denoising, Bayes denoising and VisuShrink denoising. Data used for all methods was obtained from the simulated brain MRI data available on BrainWeb. In addition to this, InhomoNET is also provided with some real brain MRI data obtained from Kaggle databases for its training since it is a deep learning solution. Each method is compared using the performance metrics SSIM, MSSSIM, and PSNR. It is found that the methods BM3D and NLM by far obtain the best results.