基于最大类间方差法与分水岭的舌图像瘀斑提取

Extraction of tongue image ecchymosis based on Otsu method and watershed

目的:研究舌图像瘀斑的自动提取算法。方法:以上海中医药大学提供的典型瘀斑舌图像为研究对象。采用了层层递进的提取思路:首先采用最大类间方差法对舌体进行粗分割,去除大部分非瘀斑区域以及与瘀斑差异较大的干扰;其次根据粗分割结果,采用形态学开闭重建预处理的分水岭算法进行细分割,得到瘀斑区域;最后对细分割结果进行后处理,去除误提取区域,优化提取结果。结果:提出了一种基于最大类间方差法与分水岭的瘀斑提取算法,使用该算法对72例典型瘀斑舌图像进行瘀斑提取,大部分结果满足瘀斑特征分析的要求。结论:该算法实现了舌图像瘀斑的自动提取,为之后瘀斑的特征分析打下良好基础。

Objective： To study the automatic extraction algorithm of tongue image ecchymosis. Methods： Typical ecchymosis tongue images provided by Shanghai University of Traditional Chinese Medicine were studied. The extraction idea of layers of progressive was adopted. First, the tongue body was roughly divided by method of maximum classes square and error（Ostu method）, and most of the non-ecchymosis areas and the the interference with the large difference from ecchymosis were removed. Secondly, according to the rough segmentation results, morphological open-close reconstruction pretreatment watershed algorithm was used for fine segmentation, then the ecchymosis areas were achieved. Finally, the result of fine segmentation was postprocessed to remove the wrong extraction area and optimized the extraction result. Results： Extraction algorithm of tongue ecchymosis based on Otsu method and watershed was put forward. The algorithm was used to extract ecchymosis from 72 typical ecchymosis tongue images, and most of the results met the requirements of the feature analysis of ecchymosis. Conclusion： The algorithm realizes the automatic extraction of tongue image ecchymosis, laying a good foundation for the characteristic analysis of ecchymosis.