A Comparison of Unsupervised Shot Classification Algorithms for News Video Segmentation
Automatic classification of shots extracted by news video plays an important role in the context of news video segmentation. In spite of the efforts of the researchers involved in this field, a definite solution for the shot classification problem does not yet exist. Moreover, the authors of each novel algorithm usually provide results supporting the claim that their method performs well on a set of news videos, without facing the problem of making a wide comparison with other algorithms in terms of key performance indexes.
In this paper, we present an experimental comparison of three shot classification algorithms. We considered only techniques that do not require the explicit definition of a model of the specific news video. In such a way the obtained performance should be quite independent of the news program’s style. For testing the selected algorithms, we built up a database significantly wider than those typically used in the field.
- 1.Furht, B., Smoliar, S.W., Zhang, H.: Video and Image Processing in Multimedia Systems. Kluwer Publishers, Boston (1996)Google Scholar
- 2.Avrithis, Y., Tsapatsoulis, N., Kollias, S.: Broadcast news parsing using visual cues: A robust face detection approach. In: Proc. IEEE Intern. Conf. on Multimedia and Expo., vol. 3, pp. 1469–1472 (2000)Google Scholar
- 3.Eickeler, S., Muller, S.: Content-based video indexing of TV broadcast news using Hidden Markov Models. In: Proc. IEEE International Conference on ASSP, pp. 2997–3000 (1999)Google Scholar
- 6.Hanjalic, A., Lagendijk, R.L., Biemond, J.: Semi-Automatic News Analysis, Indexing, and Classification System Based on Topics Preselection. In: Proc. of SPIE: Electronic Imaging: Storage and Retrieval of Image and Video Databases, San Jose, CA (1999)Google Scholar