Author : Sharu Jaiboy Mathen and Anju George

MRI images often suffer from low contrast and noise, especially in cardiac and brain imaging. Only a skilled radiologist can make an effective diagnosis and this limits its use in a wide medical network. This noise hampers further tasks such as segmentation of the important features and classification of images, 3-D image reconstruction and registration. The noise in MR images will change the value of amplitude and phase of each pixel. As a result, the visual quality gets deteriorated and perfect diagnosis of the disease becomes difficult. Enhanced processing of medical images is therefore necessary for obtaining high quality images of human tissues and organs. The SNR of the images used during quantitative analysis can be improved by using denoising methods, which improves the image quality by reducing the noise component thereby preserving all the image features. Although medical images are corrupted by different types of noises, this paper focuses on noise prominently in MR images which are Gaussian and Rician distributed. This work aims to improve the contrast and the SNR of MR images by taking into account both the homogeneous and non homogeneous nature of noise. The proposed approach provides better contrast using histogram equalization and effective denoising is achieved using Non Local Means (NLM) filter. The results are validated on simulated and real data using both visual quality assessments and performance metrics.

Keywords: Anisotropic Diffusion, Denoising, Histogram Equalization, MRI, Non Local Means filter.

Article published in International Journal of Current Engineering and Technology, Vol.4, No.6 (Dec-2014)