A color image sent to a monochrome output device must undergo a color-to-grayscale transformation. Such a transform typically retains the luminance channel or a derivative thereof. A problem with this approach is that the distinction between two different colors of similar luminance is lost. This loss can be particularly objectionable if the two colors are spatially adjacent. The paper describes a color-to-grayscale transformation technique that locally preserves distinction between adjacent colors by introducing high-frequency chrominance information into the luminance channel. This is accomplished by applying a spatial high-pass filter to the chrominance channels, weighting the output with a luminance-dependent term; and adding the result to the luminance channel. The outcome of this is that luminance variations are introduced into the image only in those regions containing high-frequency chrominance information. Regions with smoothly varying chrominance undergo little enhancement, and in particular, grayscale input is passed through without alteration. The spatially adaptive nature of the algorithm readily distinguishes it from standard approaches that apply global or pixelwise transformations. Also, preference experiments demonstrate the superior qualitative performance of the proposed approach over standard techniques.
Raja Bala, Reiner Eschbach, "Spatial Color-to-Grayscale Transform Preserving Chrominance Edge Information" in Proc. IS&T 12th Color and Imaging Conf., 2004, pp 82 - 86, https://doi.org/10.2352/CIC.2004.12.1.art00016