Skin balancing: Skin color-based calibration for portrait images to enhance the affective quality

Abstract

Because our sensitivity to human skin color leads to a precise chromatic adjustment, skin color has been considered a calibration target to enhance the quality of images that contain human faces. In this paper, we investigated the perceived quality of portrait images depending on how the target skin color is defined: measured, memory, digital, or CCT skin color variations. A user study was conducted; 24 participants assessed the quality of white-balanced portraits on five criteria: reality, naturalness, appropriateness, preference, and emotional enhancement. The results showed that the calibration using measured skin color best served the aspects of reality and naturalness. With regard to appropriateness and preference, digital skin color obtained the highest score. Also, the memory skin color was appropriate to calibrate portraits with emotional enhancement. In addition, the other two CCT target colors enhanced the affective quality of portrait images, but the effect was quite marginal. In the foregoing, labelled Skin Balance, this study proposes a set of alternative targets for skin color, a simple but efficient way of reproducing portrait images with affective enhancement.