| 英文标题 | Video motion magnification method based on SPNet |
| 作者英文名 | Zhou Fei, Du Zhenlong |
| 机构英文名 | School of Computer Science & Technology,Nanjing Tech University,Nanjing 211816,China |
| 英文摘要 | Video motion magnification technology is widely used in many fields, but because the motion of attention is too weak, it is easy to be confused with video noise, and it is easy to generate artifacts and fuzzy noise. In order to overcome the lack of detail in the traditional algorithm and the limitation of the learning-based algorithm on the magnification, this paper combined the manual design of the filter with the course learning strategy, and proposed a network structure based on the controllable pyramid decomposition, called SPNet(steerable pyramid networks). It could process a large motion range while also having a larger magnification range and could retain more details. The results of this method show that the image details are superior to the traditional algorithms and have great advantages compared with the machine learning based video motion magnification method. |
| 英文关键词 | motion magnification; curriculum learning; steerable pyramid; machine learning |