Research On Several Technologies Of Image Analysis For The Objectification Of Tongue Diagnosis

TCM (Traditional Chinese Medicine) tongue diagnosis is an unique diagnostic method, and it is an important part in TCM inspection. TCM considers that the human body is an organic whole, and every part of body relates each other closely. Because tongue has closely relation to viscera, the exuberance and decline of qi, the nature of disease and other important information of body can be gotten by observing tongue which just likes a window. The advantage of tongue diagnosis is convenience that the information of disease could be gotten quickly just by observation. Therefore, it is of great significance in guiding curing disease, expecting disease and prescribing receipt. Meanwhile, it is a rare diagnostic method which is no pain and no wound.The traditional tongue diagnosis relies on visual observing. The diagnosis results often vary from person to person. The subjectivity hampers its further development. Therefore, the research on objectification of TCM tongue diagnosis is beneficial to development of tongue diagnosis and value in application. Meanwhile, it is of great significance for TCM realizing standardization and quantification. Several key technologies of image analysis for the objectification of tongue diagnosis are studied.In the first chapter, the application background, the current situation of research and the content of this paper are expounded.In the second chapter, the acquiring environment of tongue image and the usual color space are presented. Snake model, Level Set theory, Harris corner detection algorithm, and the segmentation algorithm based on Minimal Spanning Tree are addressed.In the third chapter, the tongue body segmentation is studied. To solve the problem of weak edge, firstly, we propose a method to enhance the weak edge. Secondly, according to the tongue features and distribution features in HSV space, we realize the initialization of active contour automatically. Thirdly, in order to improve convergence at the weak edge, we construct a signed pressure force function based on regional statistics to replace the edge stopping function of Level Set model. At the end, we use a Gaussian filtering process to further regularize the level set function, and that improve the evolution efficiency of Snake.In the fourth chapter, an effective method of recognizing tooth marks is proposed. We improve the traditional Harris corner detection algorithm:(a) we propose a bilateral structure tensor to enlarge the ability to search corner.(b) a multi-scale filtering is proposed to improve accuracy of corner positioning. We propose a quantitative method to estimate the degree of tooth marks tongue.In the fourth chapter, we improve the segmentation algorithm based on minimum spanning tree (MST). (a) According the tongue feature in HSV space, we propose a method to get the tongue coat region automatically.(b) to construct a adaptive algorithm of threshold functionτ( C).(c) to propose an arithmetic operators of regional combination.(d) to improve the iterative algorithm of segmentation process.In the sixth chapter, we summarize the main content and innovative points of our paper. And the objectification research of TCM tongue diagnosis is prospected.