结合子采样和非负矩阵分解的稳健图像摘要  

Robust Image Hashing Based on Subsampling and Nonnegative Matrix Factorization

作　　者：崔得龙[1,2] 孙国玺[1,2] 邵龙秋[1,2] 

机构地区：[1]广东石油化工学院计算机与电子信息学院,广东茂名525000 [2]广东省石化装备故障诊断重点实验室,广东茂名525000

出　　处：《小型微型计算机系统》2012年第7期1629-1632,共4页Journal of Chinese Computer Systems

基　　金：广东省自然科学基金项目(8152500002000011)资助;茂名市重点科技计划项目(203624)资助

摘　　要：提出一种新颖的基于子采样和非负矩阵分解(NMF)的稳健图像摘要算法.算法首先将原始输入图像进行子采样得到一系列子图像,其次利用子采样图像的列相似性进行非负矩阵分解,最后由分解系数得到摘要序列.实验结果表明本算法对常见图像处理操作稳健,对恶意篡改敏感,同时用户密钥的使用保证了算法的安全性.A novel robustness hashing method based on subsampling and nonnegative matrix factorization is proposed in this paper.Firstly,subsampling is used to construct a subimage sequence.Then,based on the column similarity of the subimage sequence,nonnegative matrix factorization（NMF） is applied to decompose the sequence.Finally hash sequence is obtained by the factorized decomposition coefficients.Test results indicate that the proposed method is robust against common image processing,and sensitive to malignity tamper,meanwhile,the security of proposed algorithm is total depend on user key.

关 键 词：子采样 图像摘要 非负矩阵分解 归一化Hamming距离 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]