Automatic Segmentation of Sublingual Excrescences in Color Sublingual Images

Sublingual excrescence is an important diagnostic measurement in tongue diagnosis, which can assist diagnosing a variety of diseases and syndromes. This paper proposes the framework for automatic segmentation of sublingual excrescences in color sublingual images. Simply initialized GrabCut based on the visual saliency is firstly applied to segment out the dorsum of tongue where sublingual excrescence located on. And then, a stepwise method for segmenting the sublingual excrescence is proposed. Thereinto, we use the over-detection of the light-reflecting regions especially designed for the sublingual images with sublingual excrescences to eliminate the interference of regions with high brightness. Multi-threshold Otsu method is then applied to coarsely segment the image of the dorsum of tongue, and obtain the candidate excrescence regions. Finally, the fact that the protuberance of sublingual excrescence always produces shadows in its neighborhood help extract the final contour of sublingual excrescences.