Abstract
In this paper, we describe our experiments for TRECVID 2005 for the shot boundary detection task. Our approach is based on a multi-step comparison of the video frames. By measuring the difference between frames at varying distances apart, a distance map is generated and used to determine the existence of a transition and its type. The contents of the frames in a video shot are similar. If the difference between two consecutive frames is relatively large, a cut should happen. During a gradual transition, the difference between two consecutive frames is relatively small; therefore differences between the more distant frames are needed. While the comparative step size or the distance between two frames is equal to or larger than the length of transition, the difference between the frames during the transition will be much larger than that within the same video shot. In a distance map, a cut will appear as a triangle, a flash as two straight lines, and a gradual transition as a trapezoid. Based on the distance map, the different transitions can be detected and classified easily.
Original language | English |
---|---|
Title of host publication | 2005 TREC Video Retrieval Evaluation Notebook Papers |
Publisher | National Institute of Standards and Technology |
Publication status | Published - 1 Jan 2005 |
Event | TREC Video Retrieval Evaluation, TRECVID 2005 - Gaithersburg, MD, United States Duration: 14 Nov 2005 → 15 Nov 2005 |
Conference
Conference | TREC Video Retrieval Evaluation, TRECVID 2005 |
---|---|
Country/Territory | United States |
City | Gaithersburg, MD |
Period | 14/11/05 → 15/11/05 |
Keywords
- Cut detection
- Gradual transitions
- Shot boundary detection
- Video analysis
- Video partitioning
ASJC Scopus subject areas
- Computer Graphics and Computer-Aided Design
- Computer Vision and Pattern Recognition
- Human-Computer Interaction
- Software