The Extraction And Research Of Tongue Image In Traditional Chinese Medicine

Tongue diagnosis is one of the essential methods in traditional Chinese diagnosis. It plays an important role for Traditional Chinese Medicine clinical practice. However,there are limitations in traditional ways of tongue diagnosis in that it specializes on the personal clinical practices of the doctor depending on HSI subjectivity. It is easy to be influenced by the environment and visual judgment leading.The diagnostic experience cannot be completely saved scientifically in traditional tongue diagnosis,which is bad for the fluency in diagnostic techniques the medicine HSItory for the patients. Therefore,the only way to develop Chinese tongue diagnosis and improve its theoretical basis is to analyze the tongue image applying technologies of computer and image processing to make it standard, objective and quantitative.This thesis makes a thorough analysis of each of the three links in Chinese tongue diagnosis-the segmentation of the tongue body,separation of tongue and fur and the extraction and recognition of the features of tongue body. The samples in tHSI thesis consisted of fifty images from the clinical corpus of tongue image of Dr. Lowrander under the software programming of Matlab. In the part of the tongue region segmentation,using color space model conversion method to extract tongue contours,and coming up with the improved Snake model segmentation method for segmentation and astringe the edge of tongue,the results showed that HSV space model has an important role in the contour extraction process, the improved Snake method can cut apart tongue region accurately and solve the problem which exist in Traditional Snake methods. In the parts of separating moss and quality,proposed the method of dynamic threshold for multicolor channels and reached the ideal segmentation and speed. In the classification of the color feature,this thesis applied the various features of the color space model for extracting feature of tongue and proposed an improved algorithm for support vector machine for dividing multi classification, experimental results show that the algorithm can improve the recognition rate and does not affect the processing speed, better suited to solve the identification problem tongue color and the color of moss. The correct recognition rate of tongue nature is over86%;the correct recognition rate of tongue coating is over94%;tHSI method can fulfill the requirements in disease analysis of doctors or researchers.