Research On Key Techniques Of Digital Tongue Diagnosis Software Based On Clinic Tongue Images

Tongue Diagnosis is a major content of Chinese Traditional Medicine, and is the most characteristic diagnosis method. Tongue image can reflect human body's physiological functions and pathological feature keenly, thus has an important application value. With the development of computer technology, and based on a digital image processing knowledge, research on tongue diagnosis is undergoing brand new trial.This thesis focus on some new thoughts about separation of tongue and fur, color recognition, fissure detection, and also introduction and summary of predecessors'works.First of all, this article introduces tongue image capturing procedure, including standard capturing hardware choice, the capturing operation and tongue image preprocessing. Then it gives several key points of tongue extraction, which means to get the tongue's major part from original image, and region division, separation of tongue and fur. It enumerates several ways to do the separation in details, compare each method with others and lists final comparison results. Based on this separation result, this article explores further with color analysis, fissure detection and analysis.With the color analysis, based on frequently used color model, it analyses their effectiveness and lists calculation results, and details analysis with Lab color model.With the fissure detection, it introduces an algorithm which has a creative perspective by using comprises detector, extractor and analyzer to measure fissure quantitatively.The main contents and innovation of this thesis mainly contains the following there parts:(1) Separation of tongue and fur. Based upon colleague's work of tongue extraction, this paper enumerates the following method: thesholding, auto thesholding, fuzzy clustering, region splitting, region splitting and merging, and compares each method with each other, and discusses each method's practicability.(2) Color analysis of tongue and fur. Based on frequently used color model, this paper gives the subjection index formula, which is used to distinguish similarity to different tongue and fur color categories.(3) Fissure Detection. This paper introduces an algorithm which uses a fast detector to judge if there is fissure on the tongue or not first, for those images which have fissure, a modified region growing operator based on Confidence Connected could extract fissure region out, finally, an analyzer calculates FVI (Fissure Visibility Index) and FDI (Fissure Depth Index) to presents fissure properties.