Automatic Tongue Feature Extraction

• Main Authors: Yu-Cheng Hsu, 許育誠
• Other Authors: John Y. Chiang
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/28453769138666871838

碩士 === 國立中山大學 === 資訊工程學系研究所 === 98 === In recent years, Chinese medicine in the medical profession in the West triggered a wave of new wave. Chinese medicine is based on four examinations which are listening and smelling examination, inquiry, and palpation to diagnose the patient. Tongue diagnosis is also the first of four diagnostic. The result of tongue diagnosis is based on features of tongue which are diagnosed by doctor. Observation of the tongue focuses on the tongue phenomenon which is structured by the shape of the tongue, and the substance of the tongue, and the coating of the tongue. Pathology of the tongue-shaped includes the medium, fat, lean and crooked, etc. In Pathology of the tongue-substanced, tongue color includes pale, closed to pale, reddish, red, dark red, dark purple, also have some features about ecchymosis, breaken line, tooth mark, and red dot. Pathology for the coating of the tongue includes white, yellow, black, greasy, thick, thin, peeling, or no, etc. Clinically, doctors mostly rely on their own knowledge and experience when determining major lesions of a patient by observing the coloration, overall modalities, and volume of salivary on different parts of the tongue. As a result, the diagnosis tends to be limited by knowledge, experience, train of thought, and diagnostic techniques.The subjective determination is likely to be affected by the doctor’s color sensitivity and interpretation.Different doctors may come to drastically different judgments on the same tongue presentation with little overlap. Therefore, it is important to develop scientific methods that can help doctors diagnose based on standardized differentiation procedures and render reliable diagnoses in order to enhance the clinical application value of Chinese Medicine.The computerized automatic capture of characteristics shown on the images of the surface.At first, the captured image achieves brightness and color correction by brightness calibration and color calibration. Then, the original tongue images go through HSI color space conversion, detection of the control points inside and outside the surface of the tongue, curve smoothness modification and active contour model to capture images of the tongue. After that, the tongue shape, tongue fur, tongue body, and body fluid are captured from the image of the tongue.