基于变换域和形状特征的感知图像哈希算法  

PERCEPTUAL IMAGE HASHING ALGORITHM BASED ON TRANSFORM DOMAIN AND SHAPE FEATURE

作　　者：周晓炜 赵琰 Zhou Xiaowei;Zhao Yan(College of Electronics&Information Engineering,Shanghai University of Electric Power,Shanghai 200090,China;Guangxi Key Lab of Multi-source Information Mining&Security,Guangxi Normal University,Guilin 541004,Guangxi,China)

机构地区：[1]上海电力大学电子与信息工程学院,上海200090 [2]广西师范大学广西多源信息挖掘与安全重点实验室,广西桂林541004

出　　处：《计算机应用与软件》2021年第5期218-224,共7页Computer Applications and Software

基　　金：国家自然科学基金项目(61802250);广西多源信息挖掘与安全重点实验室开放基金项目(MIMS18-04)。

摘　　要：提出一种基于非下采样轮廓波变换(NSCT)分解和Zernike矩的图像哈希算法。通过预处理操作后的彩色RGB图像转换到YCbCr空间;分别对每个通道进行一级NSCT分解得到低频图像和高频纹理图像,对Y通道的高频图像使用Canny算子提取边缘后计算Zernike矩作为中间哈希序列H_(1);将3个通道得到的低频图像分块求取6个统计特征得到18个特征向量,使用PCA降维再进行压缩得到中间哈希序列H_(2);结合高频哈希序列H_(1)和低频哈希序列H_(2)作为最终哈希h。实验结果表明,该算法比几种对比算法具有更好的区分性和更高的效率。This paper proposes an image hashing algorithm based on NSCT decomposition and Zernike moment.It converted the color RGB image into YCbCr space after preprocessing,and then the primary NSCT respectively for each channel low frequency and high frequency texture images were decomposed images.The Canny operator was used to extract the edge of the high frequency image of Y channel,and the Zernike moment was calculated as the middle hash sequence H_(1).The low-frequency images obtained from three channels were divided into six statistical features to get 18 feature vectors.PCA was used to reduce the dimension and then compressed to get the middle hash sequence H_(2).The high-frequency hash sequence H_(1) and the low-frequency hash sequence H_(2) were used as the final hash H.The experimental results show that the proposed algorithm has better discrimination and higher efficiency than the comparison algorithm.

关 键 词：图像哈希 NSCT分解 ZERNIKE矩 统计特征 PCA 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]