Dense long-term motion estimation via statistical multi-step flow

We present statistical multi-step flow, a new approach for dense motion estimation in long video sequences. Towards this goal, we propose a two-step framework including an initial dense motion candidates generation and a new iterative motion refinement stage. The first step performs a combinatorial integration of elementary optical flows combined with a statistical candidate displacement fields selection and focuses especially on reducing motion inconsistency. In the second step, the initial estimates are iteratively refined considering several motion candidates including candidates obtained from neighboring frames. For this refinement task, we introduce a new energy formulation which relies on strong temporal smoothness constraints. Experiments compare the proposed statistical multi-step flow approach to state-of-the-art methods through both quantitative assessment using the Flag benchmark dataset and qualitative assessment in the context of video editing.