3D MR image denoising using higher order kernel regression

Abstract

Noise removal from images is among the challenging processes for researches. Image denoising is a crucial step to improve 3D image conspicuity and to enhance the performance of all the processing needs of quantitative image analysis. Magnetic Resonance (MR) imaging has an increasing importance in the field of medical diagnosis. MR 3D image de-noising has two features (i) tri-dimensional structure of images and (ii) the nature of the noise, which are Rician & Gaussian. Kernel regression is one of 3D non-parametric noise level estimation technique which is effective than other denoising experimental filters. The proposed Fourth order Kernel Regression (FKR) algorithm builds an efficient and robust estimator and improves the accuracy of noise and it further improves the finer estimations of pixel value and its gradients. Experimental results demonstrate positively by achieving better performance, with respect to other de-noising filters.