摘要
同图复制伪造是最为常见的图像造假类型之一。针对同图复制伪造检测过程中容易产生大量的误匹配点对这一问题,应用RANSAC算法消除原始匹配结果中的误匹配点对,并生成原始区域和复制伪造区域之间的仿射变换矩阵,进而确定原始窗口和复制伪造窗口。在窗口区域内重复块匹配和RANSAC算法,以求发现更多的正常匹配点对,提高复制伪造区域定位精度。实验表明,RANSAC算法可以有效用于常见的块匹配算法和特征点匹配算法,对缩放、旋转等不同类型的复制伪造都具有很好的效果,对不同图像特征具有很好的适应性。针对85幅不同复制伪造类型图像的总体识别率达到92.9%。
Copy-move forgery is one of the most common types of image forgeries. In order to solve the problem of mismatched points during the copy-move detection procedure,this paper eliminated the mismatched points with RANSAC algorithm,and generated affine transformation matrix between the original areas and copy-move forged areas. With the affine transformation matrix,it located the original window and the forged window. It repeated the block matching and RANSAC algorithms inside the windows to find much more matched points. Experimental results show that RANSAC is effective for block-based match algorithms and key-point based algorithms. RANSAC is available under various copy-move forgery types such as rotation and scaling. RANSAC also shows wide suitability for different image features. Its overall accuracy achieves to 92. 9% in the 85tested images with various copy-move forgery types.
出处
《计算机应用研究》
CSCD
北大核心
2014年第7期2209-2212,共4页
Application Research of Computers
基金
国家自然科学基金青年基金资助项目(61201429)
关键词
图像取证
同图复制
仿射变换
块匹配算法
image forensics
copy-move forgery
affine transformation
block matching algorithm