For the past two years several image and video super resolution methods have been proposed that are based on convolutional neural networks. This work will be more focused on the super resolution algorithms for video, but some algorithms for images will also be considered. All considered super resolution methods will be objectively evaluated and the algorithm with best performance/quality tradeoff will be implemented as a FFmpeg’ super resolution filter.

Organization

Student

Sergey Lavrushkin

Mentors

  • Pedro Souza
close

2018