KSUMM: A Compressed Domain Technique for Video Summarization Using Partial Decoding of Videos
Generally, the videos are encoded before storing or transmitting. Traditional video processing techniques are compute intensive as they require decoding of the video before processing it. The compressed domain processing of video is an alternative approach where computational overhead is less because a partial decoding is sufficient for many applications. This paper proposes a video summarization technique, KSUMM, that works in the compressed domain. Based on the features extracted from just the I-frames of the video, frames are classified into a predefined number of classes using K-means clustering. Then, the frame which is located at the border of a class in the sequential order is selected to be included in the summary. The length of the summary video can be customized by varying the number of classes during clustering. The quality of the summary was evaluated using Mean Opinion Scores method and the result shows a good Quality of Experience.
KeywordsVideo summarization Machine learning Video abstraction Compressed video processing
- 3.Kiani, V., Pourreza, H.R.: Flexible soccer video summarization in compressed domain. In: Proceedings of 3rd IEEE International Conference on Computer and Knowledge Engineering, pp. 213–218 (2013)Google Scholar
- 5.Almeida, J., Torres, R.D.S., Leite, N.J.: Rapid video summarization on compressed video. In: Proceedings of IEEE International Symposium on Multimedia, pp. 113–120 (2010)Google Scholar
- 6.Yu, J.C.S., Kankanhalli, M.S., Mulhen, P.: Semantic video summarization in compressed domain MPEG video. In: IEEE International Conference on Multimedia and Expo, Baltimore, MA, USA, 6–9 July, pp. 329–332 (2003)Google Scholar
- 14.Oh, S., et al.: A large-scale benchmark dataset for event recognition in surveillance video. In: IEEE conference on Computer Vision and Pattern Recognition (CVPR), pp. 3153–3160 (2011)Google Scholar