An effective vector filter for impulse noise reduction based on adaptive quaternion color distance mechanism

Abstract The measure of color distances plays an important role in color image processing. An effective color distance method is based on quaternion representation, which computes a color distance by weighted average of the distances of luminance and quaternion chromaticity. However, the mechanism of assigning fixed weights to luminance and chromaticity distances cannot always effectively measure color distances, since in a color image chromaticity can change significantly. To address this issue, this paper proposes an adaptive weighted quaternion color distance method. Based on the new color distance measure, the robust outlyingness ratio and local reachability density, which are defined in grayscale images, are extended to color images to implement a coarse-to-fine color noise detection operator. In noise filtering, a weighted vector median filter is employed to restore the pixels judged as noisy. Experimental results exhibit the validity of the proposed method by showing better performance in terms of both objective criteria and visual effect, compared to other widely-used color image filtering methods.