Super-resolution encompasses a range of methods to enhance the quality of videos or images by increasing their resolution. Terms like "upscale," "upsize," "up-convert," and "uprez" are used within image processing and video editing to describe this resolution enhancement. The core idea behind most super-resolution techniques involves utilizing information from multiple distinct images to create a single enlarged image. These algorithms aim to extract intricate details from each image in a sequence, enabling the reconstruction of additional frames. This approach, using multiple frames, sets super-resolution techniques apart from more basic single-frame upscaling methods that strive to create artificial details.
SRGAN is a generative adversarial network for single image super-resolution. It uses a perceptual loss function which consists of an adversarial loss and a content loss.
