Enhanced Non Local Means Filter to Denoise MR Brain Images by Using Mean Absolute Deviation Error Measure

Abstract

 Abstract 
Now-a-days, Magnetic resonance (MR) images are using for diagnosing and treatment planning. Though, MR images are 
often contaminated by several kinds of noises such as Rician and Gaussian noises which are produced by the random thermal 
motion of electronic components. It diminishes the quality and trustworthiness of the images. This paper proposes an 
Enhanced Non Local Means Filter (ENLMF) to denoise Rician type of noise from MR images. Human brain is a complex 
structure that contains millions of neurons which makes acute changes in intensity of pixels. Therefore, the earlier model of 
NLMF might be improved to remove Rician noise from brain MR image. This study, implemented Dynamic window to 
enhance the noise removal process which utilizes Mean Absolute Deviation Error (MADE) to determine the window size. The 
method implemented in publically available noise introduced data set.  The results also compared with the existing NLMF by 
visual and quantitative measures such as Peak Signal to Noise Ratio (PSNR) and Structured Similarity Index (SSIM) 
measure. The experimental results show that the proposed method provides 9% to 10% higher result than the existing NLMF