Base On Tensor Video Retrieval Research And Application

With the popular of the network and the rapid development of multimedia technology, the number of video growths at the level of exponential on the network. Because videos contains extremely abundant information, the diversity of videos information makes memory storage, copy retrieved and copyright protection become very difficult. In some extent,watermark technology can solve the above problems, but it is easily damaged and disturbed, so content-based video copy detection(CBCD) has become the hot spot. By summarizing the mainstream algorithm of every parts in the existing video copy detection framework, two novel copy video retrieval algorithms are proposed in this paper. In the paper, research work is as follows:(1)To weaken the influence of unrepresentative frames and time complexity on detecting the copy of the video on the network, a quick and efficient variable step length algorithm is proposed to select the most representative key frames. According to the characteristics of the video continuous change, the highly similar adjacent frames can replace the corresponding video between them. Firstly, the algorithm selects core region and less affected edge region of key frame and allocate different areas with different weights so as to use the Ordinal Measures(OM) transformation distance measurement to judge the similarity between two frames. Then using the sliding window to find the most similar match. The experimental results on Actual network and MUSCLE-VCD-2007 dataset show the proposed algorithm has better performance compared with the existing algorithms of copy detection feature measure.(2)This paper proposes a network video copy detection method based on tensor. On the basis of variable step key frame extraction algorithm, we use OM tensor to represent each key frame. Each key frame can be segmented into three channels(RGB) and each channel is represented by a 3 * 3 matrix. Then it uses tensor rectangle matching to accelerate the similarity calculation of each key frame. Finally sliding window is used to find the most similar match, so as to detect the query copies of video clips. The experimental results show that this method can quickly and effectively realize copy detection.