Simultaneous video synchronization and rare event detection via Cross-Entropy Monte Carlo optimization

We propose a novel approach for synchronizing multiple videos and simultaneously detecting rare events in these videos. Unlike conventional methods which deal with video synchronization and rare event detection separately, we cast these problems into an unified energy minimization framework and present a Cross-Entropy Monte Carlo (CEMC) based method to solve this problem. In our framework, rare event detection results are utilized to improve the accuracy of video synchronization. Reversely, video synchronization results are employed to efficiently detect rare events in multiple videos. Our experimental results show that our approach can accurately synchronize videos even when there is repetition of a same motion and arbitrary large time-shift between videos. Moreover, the experiments also demonstrate that our approach is advantageous in the detection of rare events in multiple videos, simultaneously, without any process of modeling or training.