Delaunay三角剖分下的视频指纹算法

Video Fingerprinting Algorithm Based on Delaunay Triangulation

视频指纹技术在解决互联网视频版权保护、重复视频检索、监管视频非法拷贝等问题上有广泛应用,因此,本文提出一种Delaunay三角剖分下的视频指纹算法.该算法首先将视频帧分块,然后将每块看作视频帧的一个特征点,每块的灰度均值视为对应特征点的特征描述值,其次对特征点的集合进行Delaunay三角剖分,最后对三角剖分后得到的各个四面体求四个顶点的特征值均值作为视频指纹,该方法得到的视频指纹既包含空域信息又包含时域信息.实验结果表明,与传统算法相比,该算法具有更优的鲁棒性.

Video fingerprinting technology is widely used in Internet video copyright protection,near-duplicate video retrieval,illegal video copy and etc.. In this paper,a video fingerprinting algorithm based on 3 D Delaunay is proposed. Firstly,the video frames are divided into different blocks,and each block is regarded as a feature point of the corresponding frame. Then,the Delaunay triangulation is applied to the feature points,and some tetrahedrons are obtained. Finally,the mean gray value of all vertices in each tetrahedron is taken as video fingerprints that can capture both spatial and temporal information. Experimental results show that the proposed algorithm has better robustness than traditional algorithms.