O(1) contrast preserving decolorization using linear local mapping

We propose an effective and efficient local decolorization method in this paper. It is an extension of the global decolorization method [6] which robustly reproduces visual appearance of a color image in the grayscale output. The improvement of the local extension is the effective preservation of the local color contrast which may diminish in the global method. Meanwhile the proposed local extension is efficient in that the computational complexity is O(1) for each pixel, which will be independent of the local kernel size. Quantitative evaluation among existing decolorization methods shows that our local extension performs favorable in both image quality and time cost. Meanwhile, our method can be extended into temporal domain for robust video decolorization.