Scene detection and retrieval of video using motion vector and occurrence rate of shot boundaries

To reduce computational stress on the Internet over the globe, cheap in cost and high speed processors are prior requirement to process on a gigantic number of videos. Searching and access of scene of video from large database, it is a user requirement to have proper indexing, browsing. Hence, it becomes a challenging problem of searching and accessing of relevant scene of end user, trivially. However, video segmentation into shot boundary detection and key frame extraction using low-and high-level features helpful to increase the realization of this task, conventionally. Beyond this, we introduce a proposed method to detect scene based on motion vector and occurrence rate of shot boundaries in video. In which, motion vectors weights and directions addresses exact scene of video with prior action on occurrence rate of shot boundaries of video, it is possible to differentiate two scenes in a movie or video. This method works empirically, and it also assigns label to the particular scene as per defined classes to represent the theme of the story, automatically. Initially, obtain the most reliable solution first and exploit each solution in the subsequent steps recursively to detect and retrieve scene in video. Expected shot boundary computation, which can be the highest probability to be the correct one and motion vector weights and directions information is also exploited in the subsequent steps. At the end comparison of performance evaluation with the existing video scene detection methods presented in literature and verified on large videos.