The AI TecoGAN is able to take low resolution images and videos and make them high definition by figuring out what's missing, However, this process is anything but simple to achieve. Just like humans can figure out what something really is from a blurry image, a computer needs to be able to do the same. Since super resolution has been an extensively studied field in AI, how does TecoGAN measure up to DUF, RBPN, and EDVR?
The main paper, "Learning Temporal Coherence via Self-Supervision for GAN-based Video Generation" by Mengyu Chu, You Xie, Jonas Mayer, Laura Leal-Taixe, and Nils Thuerey:
Adversarial Wavelet Autoencoders by Ayush Thakur and Sairam Sundaresan:
Wavelet Turbulence by Theodore Kim, Nils Thürey, Doug James, and Markus Gross: