摘要： 传统算法处理图像复制-粘贴型篡改问题时速度较慢。为此，提出一种基于投影数据主成分分析(PCA)的图像篡改检测算法。利用分块图像的行、列投影构建图像块投影特征矩阵，通过PCA对其降维，并使用字典排序法进行排序，结合图像块偏移置信距离判断图像复制-粘贴区域，完成被动取证。实验结果表明，该算法能准确找出篡改区域，与Posucue算法相比速度较快。

关键词: 图像篡改, 图像盲取证, 投影变换, 投影特征, 主成分分析, 字典排序

Abstract: Aimming at the defects of Posucue algorithm which is time-consuming in finding copy-paste forgery in image tampering, a new algorithm based on Principal Components Analysis(PCA) of projection features is put forward. A tampered image is divided into many overlapped small image blocks. The horizontal projection and vertical projection of each image block are gained to build up a matrix. Dimensionality reduction of this matrix is completed by PCA. Then Lexicographic Sorting is applied. With a threshold distance of each neighborhood block in the dimen- sionality reduced matrix, the copy-paste area can be found. Experimental result shows that this new algorithm can find the exact copy-paste areas and be more rapid than Posucue algorithm.

Key words: image forgery, blind image forensics, projection transformation, projection feature, Principal Component Analysis(PCA), lexicographic ordering

中图分类号: 

• TN911.37