摘要
将Laplacian特征映射和扩散映射的方法应用于图像数据,通过对图像热核及扩散核相关问题的研究,提出了建立扩散形态算子及运用于图像相似性比较问题的基本框架,证明了Pyt'ev形态分析理论中的投影算子及形态相关系数是扩散形态算子及扩散形态相关系数在马赛克图像模式下的特殊情形;与此同时,还对相关算法的具体实现方法和参数选择方式进行了适当改进,使得在图像比较中的相关数据实验结果得到一定的改善.
Laplacian eigenmaps and diffusion maps are applied to image data.Based on the research on the related problems of heat kernel and diffusion kernel of image,the basic framework of establishing diffusion morphological operators and comparison of image similarity is proposed.It is proved that projection operator and morphological correlation coefficient in Pyt′ev morphological analysis theory are the special cases of diffusion morphological operator and diffusion morphological correlation coefficient in Mosaic image mode.At the same time,appropriate improvement for the algorithm implementation and parameter selection is made,so that the experimental results of the related data in the image comparison are improved.
作者
段汕
张晔
张彬彬
Duan Shan;Zhang Ye;Zhang Binbin(College of Mathematics and Statistics,South-Central University for Nationalities,Wuhan 430074,China)
出处
《中南民族大学学报(自然科学版)》
CAS
2018年第3期150-155,共6页
Journal of South-Central University for Nationalities:Natural Science Edition
基金
国家自然科学基金资助项目(61374085
11301552)
关键词
热核
扩散形态算子
扩散形态相关系数
heat kernel
diffusion morphological operator
diffusion morphological correlation coefficient