Detection of documentary scene changes by audio-visual fusion

• Main Authors: VELIVELLI, Atulya, NGO, Chong-Wah, HUANG, Thomas S.
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2003
• Subjects: Gaussian Mixture Model; Semantic Context; Visual Score; Scene Change; Shot Boundary; Computer Sciences; Graphics and Human Computer Interfaces
• Online Access: https://ink.library.smu.edu.sg/sis_research/6532; https://ink.library.smu.edu.sg/context/sis_research/article/7535/viewcontent/10.1007_3_540_45113_7.pdf
• Institution: Singapore Management University

The concept of a documentary scene was inferred from the audio-visual characteristics of certain documentary videos. It was observed that the amount of information from the visual component alone was not enough to convey a semantic context to most portions of these videos, but a joint observation of the visual component and the audio component conveyed a better semantic context. From the observations that we made on the video data, we generated an audio score and a visual score. We later generated a weighted audio-visual score within an interval and adaptively expanded or shrunk this interval until we found a local maximum score value. The video ultimately will be divided into a set of intervals that correspond to the documentary scenes in the video. After we obtained a set of documentary scenes, we made a check for any redundant detections.