基于颜色特征的同源视频复制-粘贴篡改检测

Homologous Video Copy-Move Forgery Detection Based on Color Features

本文针对同源视频的复制-粘贴篡改方式,提出了一种基于颜色特征的视频篡改检测方法,能够定位篡改位置并修复原视频.首先,计算各帧的颜色直方图,提取颜色特征并字典排序,计算邻近帧的帧间相似度获得匹配帧对;其次,针对全局颜色直方图不能表示颜色的空间分布信息,提取匹配帧对的彩色边缘特征进行复检;最后利用SIFT特征点的匹配精确定位篡改位置,并恢复原视频.实验表明,本文提供的方法在SULFA数据库视频检测上具有100%的准确率和召回率.

In view of the homologous video copy-move forgery, this paper proposes a video forgery detection way based on color features that can locate the forgery location and repair the original video. Firstly, calculating the color histogram of each frame, extracting color feature and sort in alphabetical order, computing similarity between neighboring frame for matching frame. Secondly, in view of the global color histogram unable to say the color spatial distribution of information, extracting the color edge feature of the matching frame and re-inspecting. Finally, using the SIFT feature points matching to accurately position the forgery location, and restoring the original video. The experimental results show that the provided method in the paper has 100% accuracy and recall rate on the video detection of SULFA database.