基于分层聚类的单幅图像阴影消除算法仿真  

Single Image Shadow Removal Algorithm Simulation Based on Hierarchical Clustering

作　　者：李蔚妍[1] 耿霞[1] 孙未[1] 李雨 LI Wei-yan;GENG Xia;SUN Wei;LI Yu(College of Information Science and Engineering,Shandong Agricultural University,Taian Shandong 271000,China)

机构地区：[1]山东农业大学信息科学与工程学院,山东泰安271000

出　　处：《计算机仿真》2020年第6期471-474,共4页Computer Simulation

摘　　要：针对当前单幅图像阴影消除算法中,普遍存在着图像阴影消除完成时间较长、清晰度较低、成本消耗较大等问题。为解决上述问题,提出一种基于分层聚类的单幅图像阴影消除算法。通过对单幅图像进行分析,采用稀疏AP算法对单幅图像所有信息进行粗分,根据粗分结果构建单幅图像相似度矩阵,为了提高分层聚类的性能,在采用代表点聚类算法对单幅图像相似度矩阵进行再划分,合并两个阶段划分结果,得到单幅图像所有信息的划分,完成分层聚类,利用RGB空间对单幅图像阴影进行预处理,得到单幅图像对比度增强后的灰度图像,对灰度图像进行二值化后,对单幅图像阴影中心进行定位,对定位的单幅图像阴影进行消除。实验结果表明,所提出算法图像阴影消除完成时间较短、清晰度较高、成本消耗较小。In the current algorithm,completion time of image shadow elimination is long and the definition is low.Therefore,an algorithm to eliminate the single image shadow based on hierarchical cluster was proposed.After analyzing the single image,the sparse AP algorithm was used to roughly divide all the information of the single image.Based on the result,the similarity matrix of single image was constructed.In order to improve the performance of hierarchical cluster,the representative point clustering algorithm was adopted to divide the similarity matrix of single image again.In addition,the division results of two stages were combined to obtain the division of all information of single image,so that the hierarchical cluster was completed.The single image shadow was preprocessed by RGB space to get the grayscale image of single image after enhancing the contrast.After binarizing the gray image,the shadow center of single image was located,and then the shadow of single image after the location was eliminated.Simulation results show that the proposed algorithm has shorter completion time for eliminating shadow.Meanwhile,the definition is high and the cost is low.

关 键 词：分层聚类 单幅图像 阴影消除 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]