Rough set based bilateral filter design for denoising brain MR images

Abstract

Graphical abstractDisplay Omitted HighlightsThe soft computing, in precise rough set theory (RST) based class information and edge information is used in bilateral framework for medical image denoising problem.RST based framework is proposed as prior information for denoising purpose.The proposal is able to restrict the conventional bilateral filter to over smooth the region near the boundaries.It is extension of our conference paper with wide range of noise amount and different modalities of MR images. We extended it to real human brain MR images also.Performance has been compared with state-of-the-art methods and found to be satisfactory. A study on bilateral filter for denoising reveals that more informative the filters are, better is the result expected. Moreover, getting precise information of the image with noise is a difficult task. In the current work, a rough set theory (RST) based approach is used to derive pixel level edge map and class labels which in turn are used to improve the performance of bilateral filters. RST handles the uncertainty present in the data even under noise. The basic structure of existing bilateral filter is not changed much, however, boosted up by prior information derived by rough edge map and rough class labels. The filter is extensively applied to denoise brain MR images. The results are compared with that of the state-of-the-art approaches. The experiments have been performed on two real (normal and pathological disordered) human MR databases. The performance of the proposed filter is found to be better, in terms of benchmark metrics.