A Novel Approach for Color Tongue Image Extraction Based on Random Walk Algorithm

Abstract

Tongue image extraction is a fundamental step in objective diagnoses and quantitive checking of tongues. The accuracy of tongue image extraction can directly influence the results of the succedent checking in objective diagnoses of tongues. In this paper, we improved random walk image segmentation algorithm and applied it to tongue image extraction. Firstly, we utilized toboggan algorithm which adopted new classification rules to segment initial regions. Secondly, a weighted-graph was built according to initial regions in which only those adjacent regions were connected. Thirdly, random walk algorithm was applied to make the final segmentation in which a new weight function was designed for calculating the weights between the nodes of adjacent regions. Fourthly, mathematical morphology operations, i. e. inflations and erosions, were carried out on the segmentation result of the third step in order to fill small holes on the tongue region. In the experiment, we compared our method with traditional random walk algorithm. As the experiment results show, our method achieved much better segmentation effects.