【作者】 李哲； 郑江滨； 李秀秀； 李杭；

【Author】 Li Zhe+, Zheng Jiang-bin, Li Xiu-xiu, Li Hang School of Computer Science, Northwestern Polytechnical University, Xi’an 710072, China

【机构】 西北工业大学计算机学院；

【摘要】 为实现图像内容真实性鉴别,提出一种能有效检测重采样操作的图像可信度评价方法。分析了重采样操作中的插值过程给图像带来的周期性,并将其作为图像篡改的证据。文中研究了利用图像二阶差分的DFT频谱图来检测这种周期性的方法,使用区域生长技术对可疑区域不断进行修正,直至能够精确定位重采样区域。文中定义了图像可信度的概念,避免了"非真即伪"二值判断的武断性。实验表明,该方法能有效的检测出图像重采样区域,从而对图像的可信程度进行评价。更多还原

【Abstract】 In order to distinguish the authenticity of digital images, a method of credibility evaluation for digital image was proposed that can detect the traces of re-sampling. It is this presence of periodic correlations brought by the re-sampling operation that can be served as a clue of tampering. In this paper, we present deterministic techniques to detect the periodic correlations by using the DFT of second derivative, and localize the re-sampled portion of the image by using the region growing technique. The image credibility is defined to describe the reality of image, which is acceptable than just using yes or no. Results of experiments on several image forgeries demonstrate the effectiveness of the algorithm.更多还原