Static video summarization approach using Binary Robust Invariant Scalable Keypoints

Abstract

The constant demand and generation of digital video information have recently resulted in an increase in the growth of digital video content. Due to the rapid browsing of large amounts of data, content retrieval and indexing of video require an effective and advanced analysis technique. For quickly browsing, indexing, and accessing massive video archives, video summarizing approaches have been proposed. This research presents a new binary descriptor-based method for video summarization. The proposed method extracts key points and descriptors using a Binary Robust Invariant Scalable Key point (BRISK). For matching the binary descriptors between two successive frames, we employ a Brute-force method. And keyframes are extracted from each shot as the middle frame. Experiments were carried out using open video project data sets containing videos of various genres. The Comparison of user summaries (CUS) evaluation metric is used to assess the proposed method by calculating the accuracy and error rates and comparing it to other methods. As demonstrated by the experimental results, the proposed method gives good results when compared with other methods. Keywords— Video summarization, shot boundary detection, keyframe extraction, Binary Robust Invariant Scalable Keypoints (BRISK).