Extraction of Exclusive Video Content from One Shot Video

Abstract

With the popularity of personal digital devices, the amount of home video data is growing explosively. Many videos may only contain a single shot and are very short and their contents are diverse yet related with few major subjects or events. Users often ne ed to maintain their own video clip collections captured at different locations and time. These unedited and unorganized videos bring difficulties to their management and manipulation. This video composition system is used to generate aesthetically enhanced long - shot videos from short video clips. Our proposed system is to extract the video contents about a specific topic and compose them into a virtual one - shot presentation. All input short video clips are pre - processed and converted as one - shot video. Video frames are detected and categorized by using transition clues like human, object. Human and object frames are separated by implementing a face detection algorithm for the input one - shot video. Viola Jones face detection algorithm is used for separating human and object frames. There are three ingredients in this algorithm, worki ng in concert to enable a fast and a ccurate detection. The integral image for feature computation, adaboost for feature selection and an attentional cascade for efficient computational resource allocation. Objects are then categorized using SIFT (Scale Invariant Feature Transform) and SURF ( Speed Up Robust Features) algorithm.