Video frame interpolation using deep cascaded network structure

作者：

Highlights：

• It proposes the video frame interpolation scheme based on the cascaded network.

• The combination of initialization, optical flow, and blending networks is thoroughly studied and analyzed.

• Then, each network is optimally designed to yield the best performance.

• The full analysis and comparisons of the visual and quantitative results are provided.

• The proposed method greatly improves the qualitative and quantitative performance.

摘要

•It proposes the video frame interpolation scheme based on the cascaded network.•The combination of initialization, optical flow, and blending networks is thoroughly studied and analyzed.•Then, each network is optimally designed to yield the best performance.•The full analysis and comparisons of the visual and quantitative results are provided.•The proposed method greatly improves the qualitative and quantitative performance.

论文关键词：Frame interpolation,Frame-rate up-conversion,Deep learning,Convolutional neural network (CNN)

论文评审过程：Received 28 June 2019, Revised 12 June 2020, Accepted 11 August 2020, Available online 17 August 2020, Version of Record 20 August 2020.

论文官网地址：https://doi.org/10.1016/j.image.2020.115982