ABSTRACT
ABSTRACT: Tongue diagnosis is one of the most important diagnosis methods in Traditional Chinese Medicine (TCM). Tongue image segmentation is a crucial step in developing automatic tongue diagnosis system. Rough localization of tongue body is a useful preprocessing step in tongue image segmentation, which can eliminate the adverse effect of strong edges from neighboring tissues such as face and lip when extracting tongue body contour. After exploring the existing tongue rough localization method based on gray projection, we propose a modified method combining clustering with gray projecting in this paper. The proposed method first conducts clustering on image pixels’ hue components in HSI (Hue, Saturation and Intensity) space to seek multiple thresholds. Then, image thresholding and morphological operations are conducted to generate a binary image, and the largest object region is taken as the initial localization result. Finally, the localization result is refined by gray projection on pixels’ red components. Experiment results on a variety of tongue images show that the proposed method significantly improves the accuracy of tongue rough localization in comparison with the existing gray projection method.
