ABSTRACT
The conventional tongue diagnosis lacks quantitative and robust diagnosis accuracy due to its high dependence on practitioners’ experience. Now, powerful computer processors have made it possible to develop automatic computer-aided tongue diagnosis system (Pang, 2004) via image processing and pattern recognition techniques. This automatic tongue diagnosis system usually first extracts tongue body in a tongue image by image segmentation technique, then calculates the features of the tongue body by feature extraction technique, and finally uses a classifier to obtain the final diagnosis result. Hence, tongue image segmentation (i.e., extracting tongue body) is a crucial step. Some researchers have presented some algorithms (Ning, 2012; Shi, 2013) to resolve this problem. However, to date, tongue image segmentation is still a
1 INTRODUCTION
Tongue diagnosis (Kirschbaum, 2000) is one of widely used diagnostic methods in Traditional Chinese Medicine (TCM) due to its virtues such as effectiveness, painlessness, simplicity and immediacy. Tongue diagnosis has a history of at least 3000 years, and its practitioners have accumulated very rich clinical experiences on drawing physiological and pathological information according to the features of tongue body, such as color, texture, shape and coating. Eight principles of tongue diagnosis reveal that different sub-regions of tongue body can reflect health statuses of different human organs such as heart, lung, spleen and stomach. For example, tongue’s appearance is the most useful gauge for monitoring the improvement or deterioration of a patient’s health status. The color and texture features of tongue coating, which are called TCM syndromes, often reflect many diseases and human health conditions such as inflammation and infection (Hsu, 2003).
