摘要
本文将主成分分析(PCA)应用于图像哈希,设计基于特征距离的感知哈希算法。该算法从规范化图像中构造适合于数据降维的二次图像,接着对其进行PCA处理,用PCA降维特征的距离生成哈希序列。实验结果表明本文算法的接收机操作特性曲线的分类性能优于现有的3种哈希算法。大规模图像库的拷贝检测显示,本文算法有较好的拷贝检测性能。
Principal component analysis（PCA）is applied to image hashing algorithm and a perceptual hashing algorithm based on characteristic distance is proposed.In the proposed algorithm,a secondary image suitable for data dimension reduction is first constructed from the normalized image.Then,PCA is used to process secondary image,and the distance between PCA features is finally exploited to form image hash.Experimental results illustrate that classification performance of the proposed algorithm measured with receiver operating characteristics（ROC）curve is better than those of three existing hashing algorithms.Copy detection under a large-scale image database shows that the proposed algorithm has good performance in detecting image copies.
出处
《广西师范大学学报（自然科学版）》
CAS
北大核心
2016年第4期9-18,共10页
Journal of Guangxi Normal University:Natural Science Edition
基金
国家自然科学基金资助项目(61300109
61363034
61562007)
广西自然科学基金资助项目(2015GXNSFDA139040)
广西"八桂学者"工程专项经费资助项目
广西高等学校优秀中青年骨干教师培养工程资助项目(GXGQ012013059)
广西多源信息挖掘与安全重点实验室系统性研究基金资助项目(15-A-02-02
14-A-02-02
13-A-03-01)
关键词
感知哈希函数
主成分分析
数据降维
图像拷贝检测
图像检索
perceptual hash function
principal component analysis（PCA）
data dimension reduction
image copy detection
image retrieval
