This paper focuses on color image filtering, where input multichannel samples determined by a 3×3 window are adaptive directional processed in accordance with local image statistics. Thus, to form an estimate, two smoothing characteristics provided by identity filter (IF) and basic vector directional filter (BVDF) can be used. The adaptive alternation of these smoothing characteristics is based on detection criteria, where besides IF and BVDF all another outputs of 3×3 centerweighted vector directional filters (CWVDF) (a class of CWVDF includes both IF and BVDF as special cases with minimal and maximal amount of smoothing) play important role for the determination of corrupted samples. By consideration this determination, only affected multichannel samples are estimated by BVDF, whereas noise free samples are passed through IF.The proposed method extends possibilities of directional processing and represents excellent tool for impulse noise suppression with simultaneous preservation of color chromaticity, too.
Rastislav Lukac, "Adaptive Impulse Noise Filtering by Using Center-Weighted Directional Information" in Proc. IS&T CGIV 2002 First European Conf. on Colour in Graphics, Imaging, and Vision, 2002, pp 86 - 89, https://doi.org/10.2352/CGIV.2002.1.1.art00020