Video Copy Detection Research Based On The Content

The popularity of the Internet, a variety of compression technology andhigh-capacity storage technology make the spread, production, storage of thevideo more easily. Rapid growth in the number of video generates a lot ofapproximation repeat video. A large number of near duplicate video make thecopyright protection of the video very difficult, so the video copy detectiontechnology emerges.In recent years, many scholars engage in the study of video copyrightprotection technology. There are two main methods of video copy detectiontechnology, one is based on content and the other is based on digitalwatermarking technology. In digital watermarking technology, the videos needto be embedded the watermark protection before the video released, embeddeddigital watermark in the video can easily be destroyed and lose copyrightprotection, so the digital watermark to still exist some limitations in practice.The method based on content without any embedded information into the video,can directly extract robust features from video content itself, so it is convenientand the detection result shows better. The detection process is divided into thefollowing four main steps: video preprocessing, creating hashing index, feature extraction and feature matching.The key problem of video copyright protection is not only the low speed ofretrieval which is leaded by a large amount of video data, but also the lowaccuracy of retrieval which is caused by video crop, rotate, convert other videoencoding format copy attack. So, looking for good robustness and highdimensional indexing algorithm become the focuses of video copyrightprotection.In this paper, a video copy detection method based on Spherical Hashingand improved gray difference algorithm is proposed. First, the traditionalmethod of extracting key frames based on gray difference algorithm is improved.To reduce the cumulative errors and improves the accuracy of video retrieval,the method of extracting key frames based on combining the gray differencewith gray difference accumulation is used. Second, the SIFT feature of keyframes is extracted and in order to improve the retrieval speed, the famousSpherical Hashing algorithm is used to create hashing index for SIFT featurewhich changes the128high-dimensional vectors into binary hashing indexes.Experimental results show that this method not only improves the retrievalaccuracy and the retrieval speed, but also ensures a high recall rate.Experiments show that comparing with the only ordinal measure method,the method based on the GIST feature, the method based on SIFT feature andthe method on ORB feature, the proposed method keep a high recall rate andhave good precision at the same time, it also greatly improve the retrieval time. In the last, a video copy detection system based on spherical hashing andimproved gray difference algorithm is designed and implemented which is usingVC technology, the system has the advantages of portability and scalability.