Abstract: Aiming at the effect of the threshold of digital video’s authenticity by the correlation analysis, this paper proposes a digital video forgery detection method based on digital video pattern noise clustering analysis. Firstly, the video frame noise is extracted by filter, and then the pattern noise statistical properties are used as samples. Clustering analysis is used to detect low density region objects. These low density region objects are considered as tampered frames. Experimental results show that this method can effectively identify tampered video frame.

Key words: pattern noise, data mining, clustering, digital image forensics

摘要： 利用模式噪声相关性分析视频篡改取证的方法存在阈值影响的问题, 提出一种基于模式噪声聚类分析的篡改检测方法。首先用滤波器提取视频帧噪声, 然后以这些视频帧噪声的统计特性作为样本, 用基于密度的聚类算法得到低密度区域对象, 检测出被篡改的帧。实验结果表明, 该算法能够有效鉴定视频是否被篡改。

关键词: 模式噪声, 数据挖掘, 聚类, 数字图像取证