Drawing Motion without Understanding It

We introduce a novel technique for summarizing a short video to a single image, by augmenting the last frame of the video with comic-like lines behind moving objects (speedlines), with the goal of conveying their motion. Compared to existing literature, our approach is novel as we do not attempt to track moving objects nor to attain any high-level understanding of the scene: our technique is therefore very general and able to handle long, complex, or articulated motions. We only require that a reasonably correct foreground mask can be computed in each of the input frames, by means of background subtraction. Speedlines are then progressively built through low-level manipulation of such masks. We explore application scenarios in diverse fields and provide examples and experimental results.