Video Summarization via Segments Summary Graphs

Mahmut Demir, H. Isil Bozma; Proceedings of the IEEE International Conference on Computer Vision (ICCV) Workshops, 2015, pp. 19-25

In this paper we propose a novel approach to video summarization that is based on the coherency analysis of segmented video frames as represented by region adjacency graphs. Similar segments across consecutive region adjacency graphs are matched and tracked using an efficient graph matching technique. Shot boundaries are detected based on a coherency score that measures the appearances and disappearances of tracked segments. As such, it is possible to form a compact representation of each detected shot based on prevalent segmented regions and their relations - referred to as the 'Segments summary graphs'. Furthermore, the segments summary graph is amenable for further semantic analysis and understanding of the scene. Experiments on benchmark datasets demonstrate that our method outperforms the state of the art summarization approaches.