Research On Video Fingerprint Algorithm Based On Salient Region

With the development of network technology, video has become the main form of information carrier because of intuitive information acquisition and rich content, etc. However, due to the openness of the internet and convenient access to video processing tool, network video has faced many potential security risks. In video surveillance, the process of capture, transmission and storage of surveillance video is suffering from the threat of tampering with video content, which directly affects the reliability of surveillance video as a basis for judicial judgment. Video fingerprint technology which is a video processing technology is developed in such a situation. Video fingerprint provides a highly efficient and reliable means for the security monitoring of network video and surveillance video.Video fingerprint technology has formed a lot of valuable researches in the field of video copyright identification and content identification. But these researches mainly focus on accuracy and robustness of video fingerprint extraction. Research about the accurate, robust and real-time video fingerprint technology is less. Therefore, this paper studies the accurate, robust and real-time fingerprint extraction algorithm, namely video fingerprint algorithm based on salient region. This paper analyzes the development trend of the video fingerprint technology, and carries out related research work, mainly as follows:(1) Introducing the basic theory of video fingerprint, domestic and foreign research present situation, and comparative analysis of classic video fingerprint extraction algorithm. Summarizing the key problems in current video fingerprint extraction. Introducing the theoretical background of image saliency detection and analysis of the multiple typical saliency detection algorithms.(2) In view of the specific application of video monitor, a new video fingerprint extraction framework is introduced. Firstly, the saliency map is formed. Secondly, the video fingerprint is extracted based on the salient region of the saliency map. The proposed framework has a positive effect on the accuracy, robustness and real-time of the video fingerprint extraction.(3) FT(Frequency-tuned) algorithm is applied as image saliency detection. However, if the salient regions comprise more than half the pixels of the image, or if the background is complex, the background gets highlighted instead of the salient object. It is to say that the background will be significantly processed. To solve the above problems, this paper presents a semi-global FT algorithm based on boundary prior, which solves the defects of FT algorithm to some extent.(4) Video fingerprint is formed by extracting the texture features of WLD in significant areas. In this paper, an improved WLD feature extraction algorithm based on 3D space time domain is proposed. Extending 2D WLD to 3D WLD, which enables the video fingerprint to contain not only the spatial information but also the time information.(5) The feasibility of the framework of the video fingerprint extraction is verified, and the performance of the algorithm is analyzed. At the same time, a real-time video content tampering detection system is designed.