Super resolution is the process of up-scaling and improving the details of an image. Currently the super resolution modules within OpenCV are based on methods such as robust regularization and optic flow estimation, while the current state-of-the-art methods are based on deep learning. I propose to add learning-based super resolution methods to OpenCV. This will allow for more accurate and faster (real-time) super resolution.

Organization

Student

Xavier Weber

Mentors

  • Yida Wang
  • Vladimir Tyan
close

2019