Automated Tongue Image Segmentation in Tongue Diagnosis

In tongue diagnosis, it is difficult to maintain patient’s tongue extended for a little bit long time, especially for pediatric patients. Fast tongue image capture and automated segmentation is the crucial procedures in the tongue diagnosis system. Its accuracy affects the subsequent automatic process directly. In the paper, an automated tongue image segmentation method was proposed. One of our main contributions is the combination of the GVF (gradient vector flow) algorithm and the ellipse shape based on the least-squares ellipse fitting algorithm as seed points; and the other is the convergence analysis for proposed algorithm combined with the transform matrix algorithm was performed. Before segmentation, a face mask was proposed based on Anil K. Jain’s theory, and the initial tongue edge was extracted with log-polar transform, as a preprocessing. The experiment results indicate that our method speeds up contour convergence, ensures correctness, reduces the system’s operating expenses, and enhances the algorithm usability. This research focused on tongue diagnosis involved in image segmentation and color correction.