Camera Motion Quantification and Alignment

We propose a method to synchronize video sequences of distinct scenes captured by cameras undergoing similar motions. For the general camera motion and 3D scene, the camera ego-motions are featured by fundamental ratios obtained from the fundamental matrices. In the case of pure translation, translational magnitude features is used. These extracted features are invariant to the camera internal parameters, and therefore can be computed without recovering camera trajectories along the image sequences. Consequently, the alignment problem reduces to matching sets of feature vectors, obtained without any knowledge of other sequences. Experimental results demonstrate the accuracy and applications of the proposed method