Tongue Image Segmentation And Feature Extraction Algorithm Based On Tongue Diagnosis In Traditional Chinese Medicine

The quantitative study of tongue diagnosis has promoted the development of tongue diagnosis in TCM,which has made the Chinese nation's experience of saving lives more effective in the past thousands of years.In this paper,the main task of quantitative study of tongue diagnosis is to complete the automatic segmentation and feature recognition of tongue by means of computer image processing algorithms.Firstly,the tongue image is divided into two types according to the distribution of hue gradient and divided separately.When the tonal gradient distribution of tongue body is uniform and the background gradient is large,the improved Otsu algorithm is used to segment the tongue,and the edge is repaired with the help of canny operator and morphological processing,and the excellent segmentation result is obtained.When the tone gradient of the tongue is unevenly distributed,The angle channel and chord length channel are constructed by using the spatial geometrical characteristics of s channel and I channel in HSI color space in three-dimensional color space,and the iterative threshold segmentation algorithm is used to segment it.Because the twovalue image of the angle channel has better effect on the lower part of the tongue region,the twovalue image of chord length channel has a better effect on the upper part of the tongue,so a new adaptive image mosaic algorithm is used to stitch the tongue body region with two two-value image segmentation effect into a complete tongue image,and carry on the operation of the connected domain and the morphological processing,the better segmentation effect of tongue is obtained.If the color characteristics of the tongue are extracted under the RGB space conditions,the eigenvalues of different pathological tongues will overlap.Therefore,the color characteristics of the tongue are defined by using the HSI space with different channel value range.Because of the geometrical characteristics of HSI space,there is no definition of H channel on the axis of the double cone,so two new channels are introduced to help classify the color characteristics of the tongue.For the identification of the tooth-mark Tongue,this paper first analyzes the roundness of the tongue,the geometrical features of aspect ratio and rectangle degree,and then aiming at the concave and convex of the edge of the tooth mark Tongue,combining the definition of roundness and rectangle degree,this paper puts forward two geometrical features of convex and convexpackage difference,and proves that the convex package degree is accurate in classifying the teethmarked tongue.