A Novel Three-phase Approach for Highly Corrupted Color Images
Ling Zhong, Yonghui Huang, Xiangqian Xiao · 2017
To reduce impulse noise in color images, a novel three-phase iterative filter is presented in this paper. The proposed filter first detects noisy components in each color channel by applying the detection scheme of the adaptive median filter (AMF). It then utilizes detected noise-free components to construct the combined gradient for each pixel, iteratively. Finally, an iterative variant of the vector median filter (VMF) is utilized to remove the rest of completely corrupted vectors. Extensive experiments show that the presented filter achieves significant improvement over other existing vector-based filters when the images are highly corrupted by impulse noise.