Automatic Image Retrieval through Video Authoring and Transition

Akila. R, Jayashree. B, Mr.P. Karthick
2014 IOSR Journal of Computer Engineering  
An integrated system for video summarization, browsing and presentation, based on large amount of personal and web video clips. Content-consistent shots are retrieved from a video pool in order to form a descriptive long-shot video automatically by video authoring and transition to present events, persons or scenic spots captured at various remote places in an informative manner. Users accessing the application have to register their information via user authentication. Recognized users records
more » ... and their personal information are secured and maintained by the administration. Input short videos are converted into frames in pre-processing where each frames are resized and merged into a single video for video categorization. Videos are categorized by using transition clues like human, object. The frames are categorized into human and nonhuman frames by using Viola-Jones algorithm where Region Of Interest is used to separate the frames. Human frames are composed with and without reference image by using Back Propagation Network with the help of Trainee Database which consists of list of human frames with different angles in a specific Group. Non Human frames are composed with and without Reference image by using Speeded up Robust Features where object and sequence matching process are performed. Object frames and related sequence frames are categorized into a separate folder and converted into separate single shot videos. The system provides an efficient video browsing mode to generate matching graph of videos.
doi:10.9790/0661-1628119125 fatcat:rwp7g6yum5f77d2byodzygqvmq