We proposed a new technique of detecting facial pattern from negative color film based on the statistical pattern recognition method. After gathering the statistics data of their lightness and chromaticities, skin color region was extracted by using binary image processing. On the basis of probability distribution of five pattern variables in the labeling image, we detected the facial patterns. This detection makes color correction of facial pattern regions possible to do by applying local color correction technique. The chromaticities of skin color in many people pictures printed by offset printing were measured for comparison with the preferred color reproduction result. Afterwards, we produced facial pattern images in many different colors and evaluated these obtained images to estimate the preferred color reproduction of facial pattern in hardcopy.
Demas Sanger, Takuya Asada, Hideaki Haneishi, Yoichi Miyake, "Facial Pattern Detection and Its Preferred Color Reproduction" in Proc. IS&T 2nd Color and Imaging Conf., 1994, pp 149 - 153, https://doi.org/10.2352/CIC.1994.2.1.art00041