为了从原始舌图像中精确分割出舌体以实现中医舌诊的客观化,提出了一种基于Otsu阈值法与形态学自适应修正的舌像分割方法.该方法提取原始舌像的RGB色彩空间与HSV色彩空间中G、B、V通道的信息,运用Otsu阈值法对舌像进行二值化,基于先验知识对目标区域进行聚类,采用形态学运算方法,利用舌体当量直径自适应选取结构元素的大小,对聚类结果进行开启运算修正得到舌像分割结果.基于200幅不同形态、不同舌色类别原始舌像的实验结果表明,该方法能稳定精确地分割出舌体,与传统的Otsu方法及灰度投影-Otsu方法相比,准确率有了很大的提升.%To accurately separate the tongue body from original tongue images to realize the characterization of tongue diagnosis in traditional Chinese medicine, a new tongue image separation method based on the Otsu threshold method and morphology adaptive correction is put forward.The method extracts the information of the channels of G, B and V in the RGB color space and the HSV color space of original tongue images, performs binarization for tongue images with the Otsu threshold method, clusters target areas based on priori knowledge, self-adaptively selects the size of structure elements with the tongue body equivalent diameter and the morphological algorithm, and gives the clustering operation correction to obtain the tongue image separation results.The results of the experiments on 200 original tongue images with different forms and different tongue color categories show that this method can separate the tongue body stably and accurately, with greater improvements in accuracy compared with the traditional Otsu method and the gray projection-Otsu method.
展开▼
