Multi-scale adaptive factors video acceleration magnification

The subtle variations in video may contain some significant information, and the variations may be tiny motion of objects or ones superimposed on large motion. Traditional video magnification methods can only amplify the single micro-motion. The state-of-the-art video acceleration magnification method can handle with linear large motions while still amplifying small changes. However, it still introduces some artifacts or blurring. In this work, we propose a method to mitigate this issue by adopting a magnification factor adaptive strategy, which is based on analyzing the limitation of magnification factor at different scales of the complex steerable pyramid. In this new approach, magnification factors at various scales are set according to the spatial representative wavelengths of each scale. Experiments have been carried out on both synthetic and real videos. Results demonstrate that our method is effective in amplifying tiny motions superimposed on large motion, getting sharper visual quality and mitigating artifacts or blurring.