Detection of Tongue Crack Based on Distant Gradient and Prior Knowledge

Tongue crack is an important pathological feature in traditional Chinese tongue diagnosis. However, in computerized tongue diagnosis, tongue cracks are not commonly used because the tongue crack features are hard to be accurately extracted. In this paper, we propose a new scheme for tongue crack extraction. First, for the purpose of enhancement of the tongue crack regions, we get line response image using tongue image gray-level and color information as well as the pixel distant gradient. Then tongue crack regions are extracted from line response image by methods such as hysteresis thresholding algorithm and so on. A database of 286 tongue images is established and used in our test, and the experimental results demonstrate the effectiveness of our proposed tongue crack detection method.