We propose two algorithms for performing highlight elimination from color images. The first algorithm is based on analyzing the spatial distribution of a surface's color in RGB space, and the second algorithm uses mostly image space information. For the second algorithm we assume the imaged surface to be a cylinder; the first algorithm is able to eliminate highlights from images of surfaces of arbitrary shape. Advance knowledge of the intensity or direction of the illumination, or the surface's diffuse or specular reflectances is not required.
Michael Hild, "Elimination of Highlights using RGB Color Distribution and Image Position" in Proc. IS&T 4th Color and Imaging Conf., 1996, pp 205 - 208, https://doi.org/10.2352/CIC.1996.4.1.art00053