A new denoising filter for brain MR images

Abstract

A new denoising filter is proposed for human brain MR image. The proposed filter is based on the notion of existing bilateral filter whose objective is to get a noise-free smooth image, preserving edges and other features intact. We have introduced a weighing function that controls the impact of existing bilateral filter for denoising. It is conditioned by Rough Edge Map (REM) and Rough Class Label (RCL). The presence of noise makes difficult to get precise information of edge and class label. Rough Set Technique is expected to assign rough (imprecise) class label and edge label to the pixels in the given image. This function thus is expected to handle the impreciseness of edge and class label and thereby preserving these two by controlling the bilateral filter more efficiently. The filter is extensively applied on brain MR images. The current proposal is compared with some of state-of-the-art approaches using different image quality measures and found to be efficient in most of the cases.