A Motion-Driven Approach for Fine-Grained Temporal Segmentation of User-Generated Videos.

This paper presents an algorithm for the temporal segmentation of user-generated videos into visually coherent parts that correspond to individual video capturing activities. The latter include camera pan and tilt, change in focal length and camera displacement. The proposed approach identifies the aforementioned activities by extracting and evaluating the region-level spatio-temporal distribution of the optical flow over sequences of neighbouring video frames. The performance of the algorithm was evaluated with the help of a newly constructed ground-truth dataset, against several state-of-the-art techniques and variations of them. Extensive evaluation indicates the competitiveness of the proposed approach in terms of detection accuracy, and highlight its suitability for analysing large collections of data in a time-efficient manner.