Tongue image segmentation via thresholding and clustering

Abstract

Tongue diagnosis is widely used in Traditional Chinese Medicine (TCM). Tongue image segmentation plays a key role in developing an automatic tongue diagnosis system. To achieve tongue image segmentation (i.e., extraction of tongue body), we propose a method based on image thresholding and image clustering. The proposed method first finds an initial object region mainly containing the tongue body and the upper lip by transforming and thresholding on the hue component of an image in HSI color space, as well as subsequent morphological operations. Then, our method performs image clustering on the RGB components of the initial object region to seek the gap region between the tongue body root and the upper lip. Finally, the tongue body is extracted from the initial object region by removing fake tongue body regions such as the upper lip with the help of the gap region. Experiment results on a series of tongue images demonstrate the effectiveness of the proposed method.