摘要
医学图像处理过程通常包括图像预处理、特征提取、图像分类,Harris角点检测算法是常用的特征点提取算法之一。该算法适应多种变换、运算简便,在医学图像处理领域中广泛应用,但在实际应用中发现传统的Harris算法检测到的特征点数量不足且图像配准精度不高。因此提出了一种优化算法(GM-Harris算法),即采用群搜索优化算法(GSO算法)与互信息相结合的方式优化传统Harris算法的过程,并从匹配有效率与算法效率两方面对2种算法的特征点提取效果进行了定量分析。实验结果表明,与传统的Harris算法相比,GM-Harris算法不但可以获得较充足的特征点,而且还能提高图像配准的精度。
Medical image processing usually includes image preprocessing,feature extraction and image classification.Harris angular point detection algorithm,one of the commonly used feature extraction algorithm,is characterized by adaptability to various transformations and simplicity of operation.However,the number of image feature points detected by it is insufficient and the accuracy of its image registration is low.A GM-Harris algorithm which optimizes the process of Harris algorithm by combining the group search optimization(GSO)algorithm with mutual information was thus proposed in order to overcome the shortcomings of traditional Harris algorithm.The feature points detected by the two methods were quantitatively analyzed in aspects of their matching efficiency and algorithm efficiency,which showed that the number of feature points detected by GM-Harris algorithm is greater than that detected by traditional Harris algorithm and the image registration accuracy of GM-Harris algorithm is higher than that of traditional Harris algorithm.
作者
赵悟
段永璇
段会川
张睿
肖宪翠
岳媛
ZHAO Wu;DUAN Yong-xuan;DUAN Hui-chuan;ZHANG Rui;XIAO Xian-cui;Yue Yuan(Shandong Medical and Health Information Institute/Shandong No.1Medical University,Jinan 250062,Shandong Province,China;Shandong Normal University Information Science and Engineering School,Jinan 250014,Shandong Province,China)
出处
《中华医学图书情报杂志》
CAS
2019年第11期22-26,共5页
Chinese Journal of Medical Library and Information Science
基金
国家自然基金项目“网络资源协同管理与多目标虚拟网络映射问题研究”(61373149)
山东省医药卫生科技发展计划项目“基于统计学习算法的精准疾病知识库构建研究”(2018WSA18030)
山东省医学科学院面上项目“医学信息及数据服务云平台建设研究”(2016-01)。
关键词
HARRIS角点检测算法
GSO算法
互信息
匹配有效率
算法效率
Harris angular point detection algorithm
GSO algorithm
Mutual information
Matching efficiency
Algorithm efficiency