Image De-noising by an Effective SURE-Based Weighted Bilateral Filtering

Jian Ji, Sitong Li, Guo-Fei Hou, Fen Ren, Qiguang Miao

Abstract: The weighted bilateral filtering has been widely used in image de-noising, and there have been numerous improved bilateral filters. When dealing with the images with different noise levels, however, they may get de-noising results that are not robust and cannot remove salt and pepper (s&p) noise effectively. In this paper, we propose a new weighted bilateral filter, which preserves edge information and improves robustness, for image de-noising. The new filter introduces a new pseudo-median bilateral filter (NMBF) combined with a robust bilateral filter (RBF) in a weighted way for removing additive Gaussian and s&p noise. In addition, given the edge preserving, a robust estimation kernel function is applied in NMBF. Besides, to get optimal parameters, the modified filter involves the Stein’s unbiased risk estimate (SURE) method. A diversity of images polluted by various degrees of Gaussian, s&p, and mixed noise were used to evaluate the performance of this new bilateral filter.