Isolated Vector Median Filtering for Noise Reduction in Digital Color Images
Abstract
This paper deals with de-noising in digital color images. The conventionally used median filtering, the vector median filtering and the basic vector directional filtering and their extensions for image de-noising filter the vector pixels jointly in the red, green and blue intensities. Consequently the smoothing applied in one color component smears into the others due to the correlation preserving property of these filters. This could potentially lead to additional pixel corruption in good pixels. The recently proposed isolated vector minimum distance filtering deals by isolating the joint vectors in isolation and then minimising the aggregate distance between the pixels. This technique has proved to be much more effective for reducing noise in digital color images. This paper extends their work and proposes the isolated vector median filtering approach. The key idea of this paper is to isolate the joint vector pixels and then perform median filtering on only the most important pixels, i.e., pixels with the minimum aggregate distance from each other. We demonstrate the superiority of the proposed method compared against the aforementioned methods using several test images and statistical measures.