Efficient Filtering and Clustering Methods for Temporal Video Segmentation and Visual Summarization
✍ Scribed by A.Müfit Ferman; A.Murat Tekalp
- Publisher
- Elsevier Science
- Year
- 1998
- Tongue
- English
- Weight
- 700 KB
- Volume
- 9
- Category
- Article
- ISSN
- 1047-3203
No coin nor oath required. For personal study only.
✦ Synopsis
objective reference for users and applications alike in terms Automatic temporal segmentation and visual summary gen-of absolute frame numbers or time codes. The lack of such eration methods that require minimal user interaction are key well-defined generic segments is one of the reasons higher requirements in video information management systems. Cluslevel content representations (e.g., in terms of scenes of tering presents an ideal method for achieving these goals, as sequences) are difficult to obtain.
it allows direct integration of multiple information sources.
Identifying the shot boundaries within a video sequence This paper proposes a clustering-based framework to achieve expedites random nonlinear access; however, the user must these tasks automatically and with a minimum of user-defined still view a shot in its entirety to access its visual content.
parameters. The use of multiple frame difference features and
In other words, simple temporal segmentation provides short-time techniques are presented for efficient detection of only the backward/forward shuttling capability, but it does cut-type shot boundaries. Generic temporal filtering methods not enable browsing. A compact representation of the viare used to process the signals used in shot boundary detection, resulting in better suppression of false alarms. Clustering is sual content for each shot must be provided to reduce also extended to the key frame extraction problem: Color-based storage requirements and to introduce true browsing funcshot representations are provided by average and intersection tionality. Perhaps the simplest way to describe shot content histograms, which are then used in a clustering scheme to is by textual keywords. While text-based descriptions of identify reference key frames within each slot. The technique content can relay a great deal of information, their generaachieves good compaction with a minimum number of visually tion requires intense human interaction. Furthermore, connonredundant key frames.