基于一种局部图像增强和改进分水岭的舌体分割算法

A tongue segmentation algorithm based on local image enhancement and improved watershed

舌体分割是舌诊客观化中非常重要的第一步,鉴于舌体与唇色相近,故难以将二者分开。基于此,该文提出一种舌体分割算法,该算法先用改进的分水岭算法和Otsu算法获取舌体的初始区域,并且剔除其中的部分冗余信息;接着,针对初始区域中舌体和下唇区分的难题,提出一种局部图像增强(LIE)方法,先从初始区域中,然后利用局部图像中舌体与下唇细微的色差,计算出不同的图像增强系数,使增强后的局部图像中舌体和嘴唇的颜色具有明显的区别,从而将两者精确地划分。经试验证明,该舌体分割算法的评估指标FN,FP,TP分别为3.43%,2.89%和96.6%。因此,该文算法具有良好地分割效果。

Tongue segmentation is the first very important step in the objectification of tongue diagnosis.Since the color of tongue is close to that of lip,it is difficult to separate them.In view of this,a tongue segmentation algorithm is proposed in this paper.Firstly,the improved watershed algorithm and Otsu algorithm are used to obtain the initial region of the tongue and remove the redundant information.And then,in order to solve the problem of separating the lower lip and the tongue in initial region,a local image enhancement(LIE)method is proposed.The subtle color difference between tongue and lower lip in the local image is taken to calculate different image enhancement coefficients and make the colors of tongue and lip in the enhanced local image have obvious distinction,so as to precisely divide tongue and lip in the enhanced local image.Experimental results show that the evaluation indexes of FN,FP and TP in the tongue segmentation algorithm are 3.43%,2.89%and 96.6%respectively.It means this algorithm has a good segmentation effect.