Video Copy Detection Using F-Sift and Graph Based Video Sequence Matching Anju P S #1 , Soumya Varma *2 , Vince Paul #3 , Sankaranarayanan P N #4 1 Student, Computer Science and Engineering Department, Calicut University, Sahrdaya College of Engineering and technology 2 Assistant Professor, Computer and Engineering Department, Sahrdaya College of Engineering and technology 3 Head of Department, Computer and Engineering Department, Sahrdaya College of Engineering and technology 4 Assistant Professor, Computer and Engineering Department, Sahrdaya College of Engineering and technology Abstract— Content based copy detection aims to detect copies that of a given media. As with the growth of technology more and more media contents are available in the internet. This large number of copies leads to violation of digital rights. So we need an effective and efficient method to detect duplicated media contents. An auto dual threshold method is used to eliminate redundant video frames of a video segment which will reduce non necessary matching of video frames. Then used local features of F-SIFT for video content description. Flip-invariant SIFT (or F-SIFT), that preserves the original properties of SIFT while being tolerant to flip like transformations. Since matching computational cost of F-SIFT is very large, so uses an SVD-based technique to match two video frames with the SIFT point set descriptors. To obtain the video sequence matching result propose a graph- based method. It is used to convert the video sequence into identifying the longest path in the frames to identify the video matching-result with time constraint. Keywords— FSIFT Feature, graph, SIFT Feature, SVD, and graph Based matching I.INTRODUCTION With the quick growth in the internet and multimedia technology, we are able to access and store huge volumes of video data easily. That is huge volumes of videos are transmitted, searched and stored on the internet. Some statistics of the YouTube shows that, there are about 100 hrs of user generated videos are uploaded to YouTube every minute. According to BBC motion gallery, it contains over 2.5 million hours of professional video contents. Among these huge volumes of videos there exist a large numbers of duplicated and near duplicated videos. It is reported that about 27% videos in a video search results obtained from YouTube, Google & yahoo videos are duplicated or near duplicated copies of a popular version. For particular queries, the redundancy can be as high as 93%.A duplicate video means we can divide it into two Duplicate Videos and Nearly Duplicated Videos. Duplicated Video will be extracted video copies that can be easily detected. Near Duplicated video copies are transformed video clips and detection of such copies is challenging. So we can define a video copy as, it is a segment of video derived from another video usually by means of various transformations such as addition, deletion, modification and cam coding. According to the definition of video copy in TRECVID2008 tasks, A video V1, by means of various transformations such as addition, deletion, modification(of aspect, colour, contrast, encoding, and so on),cam cording, and so on, is transformed into another video V2,then video V2 is called a copy of video V1. In content based copy detection task of TRECVID 2008, 10 Transformations are defined. T1. Cam-cording; T2. Picture in picture; T3. Insertions of pattern: Different patterns are inserted randomly: captions, subtitles, logo, sliding captions; T4. Strong re-encoding; T5. Change of gamma; T6, T7. Decrease in quality: Blur, change of gamma (T5), frame dropping, contrast, compression (T4), ratio, white noise; T8, T9. Post production: Crop, Shift, Contrast, caption (text insertion), flip (vertical mirroring), Insertion of pattern (T3), Picture in picture (the original video is in the background); T10. Combination of random five transformations among all the transformations described above. Fig 1 shows image Examples for 10 transformations, Fig 1 These videos come from MUSCLE –VCD-2007 and TRECVID 2008 Anju P S et al, / (IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 6 (1) , 2015, 152-158 www.ijcsit.com 152