One-take video boundary point extraction based on frame characteristics

• Published in: 2021 International Conference on Information and Communication Technology Convergence (ICTC) pp. 524 - 526
• Main Authors: Lee, Jaewon, Jang, Dalwon, Lee, JongSeol
• Format: Conference Proceeding
• Language: English
• Published: IEEE 20.10.2021
• Subjects: Feature extraction; Filtering; Image segmentation; Information and communication technology; Mobile handsets; One-Take video splitting; Real-time systems; Real-Time Video Splitting; Shot Boundary Detection; Streaming media; Video Splitting System
• DOI: 10.1109/ICTC52510.2021.9620953

This paper deals with a system for extracting boundary points for unedited video. The term 'unedited' means that the user continuously shoots at once without stopping the shooting through a device such as a camera or mobile phone. In general, if image intrinsic pixel values or RGB characteristics are used to compare adjacent frames, it is difficult to extract a boundary point from continuous one-take video. However, the pixel mean and the standard deviation was obtained and the ratio was used as a primary filtering. Through key point extraction and comparison using SURF(Speeded-Up Robust Features), segmentation points were well extracted even in one-take videos with severe shaking. The entire system is tested with the a video shot directly through a mobile phone and a short film shot with one-take, and found that the speed of about 30fps of more was obtained, indicating real-time performance.