46 Statistical Tongue Color Distribution and Its Application

Knowledge of human tongue color and its distribution are essential for image analysis in the computerized tongue diagnosis system. Based on a large and comprehensive tongue image database, which contains over 5200 high-quality tongue images, this study aims to provide the first statistically described human tongue color distribution in CIE xyY color space and its application on color feature extraction. Representative color pixels are firstly extracted for each tongue image to reduce the total number of color pixels by the proposed blockimage boundary searching algorithm, and then they are combined together to generate the tongue color distribution. After that, one-class support vector machine (SVM) is employed to mathematically describe the boundary of this color distribution thus to separate tongue or non-tongue colors. Finally, we show the application of the proposed color distribution on color feature extraction. We believe this work is the most comprehensive and detailed study on tongue color distribution to date.