Authors : Isra`a Hadi and Hikmat Z. Neima
Abstract: Nowadays, multimedia applications are widely extended and efficient methods for video analysis are required. Video Shot Boundary Detection (SBD) is an essential step for video analysis, indexing, retrieval and summarization. SBD is the process of finding the location of boundaries between video shots. Many SBD methods have been proposed, however, most of them focus on increasing the detecting accuracy which logically leads to expensive computation time. In this study, a fast and efficient SBD method is proposed. The proposed method has two stages, namely, shot boundary identification and verification. In the first stage, input video is reduced spatially and temporally. For the spatial reduction, Extend Salient Region (ESR) method is used to reduce the process only on the important part within each frame. While Dynamic Candidate Segment Selection (DCSS) is utilized in terms of temporal reduction. In DCSS, only the frame segments that highly probable to have shot boundary is passed for the verification stage. In the verification stage, Edge Count Ratio (ECR) is applied to obtain frame local feature in order to minimize the false detected boundaries. The proposed method showed good results in regard to detecting accuracy and computation time.
Isra`a Hadi and Hikmat Z. Neima, 2017. Shot Boundary Detection Based on Dynamic Candidate Segment and Motion Estimation. Journal of Engineering and Applied Sciences, 12: 8987-8993.