Adaptive sublingual vein extraction based on Bayesian decision

A novel segmentation technique of sublingual vein extraction was proposed. The sublingual vein color model which combined the Bayesian decision with CIEYxy color space was established based on a large number of labeled sublingual images. During the segmentation process, the sublingual vein and non-sublingual vein parts were treated as independent classes, whose within-class scatter matrix and between-class scatter matrix were used to assist segmentation by defining the similarity measurement. The experiments proved that the proposed method performed very well on the segmentation of images from normal people with weak color contrast between sublingual vein and its background.