Automatic Facial Complexion Classification Based on Mixture Model

摘要

Classification of facial colors plays a vital role in Traditional Chinese Medicine (TCM), photo beautification, matching cloths and other beauty and cosmetics industry. The face color of a person is considered as a symptom to reflect the physical conditions of organs in the body. Most current methods are difficult to accurately classify the facial colors. In this paper, we propose a facial color classification method based on complexion Gaussion Mixture Model (GMM) and SVM to address this problem. Specifically, we iteratively confirm the complexion pixels belonging to the skin region based on the GMM. In the optimizing process, we extract features based on two-dimensional GMM to describe main color and minor color. Experiments are performed on our dataset with 877 face images. Experimental results demonstrate the accuracy of the proposed classification method compared with the state-of-art facial color classification method.