Super resolution aims to recover a plausible high resolution version of a low resolution image. The main techniques that were used to solve this problem were interpolation, enhancing linear methods and patch-based methods. Nowadays, state of the art in super resolution is Deep Convolution Neural Networks. These networks exploit low and high level image features to reconstruct perceptually realistic photos and videos.
Some examples of Image SuperResolution
Some examples of Video SuperResolution
The research behind
The network architecture basis for SuperResolution
- Input: Low resolution images (RGB)
- Output: Super resolution images (RGB)
- 3-layer network with one subpixel convolution network