Facial color feature extraction for disease diagnosis using non-base colors

Disease diagnosis via an individual's facial color has been practiced in oriental medicine for thousands of years. This work tackles the issue of quantitative facial color feature extraction for diagnostic purposes. Existing work in this area directly utilizes classic methods found in the standard color analysis area. We believe that these standard chromatic methods are not the most suitable for describing facial colors. Hence, we propose a color extraction method specially designed for facial diagnosis. The main contribution of this work is to define the facial base colors and to present six non-base colors (centroid colors) to represent the facial color space. First, a facial color gamut in the CIEx-y chromatic diagram is formed. Next, an area composed of base colors is defined within the color gamut. Afterwards, some clusters are located outside the base colors area. Finally, six centroids are selected to form a set of components to describe the facial color features. Disease detection is performed to compare the performance of the proposed and existing methods. Overall, the experimental results show that the proposed method outperforms the existing methods, which verifies its validity and effectiveness.