Facial Color Management for Mobile Health in the Wild

Nowadays, mobile technologies have changed the patient routine health care and management. With a large amount of mobile health applications developed, massive and valuable health data are possibly collected with a smart mobile phone in hand. Facial color images are recently proved to be available and effective for health condition diagnosis both in modern medicine and ancient medicine perspectives. One significant issue of facial color health condition diagnosis system is color management, in which its primary procedure is to obtain reliable and device-independent facial color images in the wild. The solution is known as utilizing color correction technology to recover the intrinsic color properties of facial skin. However, current color correction approaches are hard to meet the need of mobile health management in the wild, due to some limitations of precision-challenged algorithm, inconvenient color imaging device, strong scenario assumption and so forth. Therefore, in this paper, we consider several facial skin color characteristics and show that it is valuable to build facial color related correction model for facial color images in the wild. Then we propose two kinds of facial color correction strategies to realize the facial color management of mobile health in the wild. The first one is reference-based approach, and the other one is skin-based approach without requirement of colorchecker. Experimental results with qualitative and quantitative assessments on the indoor and outdoor scenarios demonstrate that the proposed reference-based approach is more outstanding than our previous method and other color constancy methods. In addition, given a facial color image only, the skin-based method can still achieve effective results compared with other color constancy methods.