Abstract
Image super-resolution reconstruction (SR) aims to use a specific algorithm to restore a low-resolution blurred image in the same scene into a high-resolution clear image. Due to its wide application value and theoretical value, image super-resolution reconstruction technology has become a research hotspot in the field of computer vision and image processing, and has attracted widespread attention from researchers. Compared with traditional methods, deep learning methods have shown better reconstruction effects in the field of image super-resolution reconstruction, and have gradually developed into the mainstream technology. Therefore, this paper classifies the image super-resolution reconstruction problem systematically according to the structure of the network model, and divides it into two categories: the super-division method based on the convolutional neural network model and the super-division method based on the generative confrontation network model. The main image super-resolution reconstruction methods are sorted out, several more important deep learning super-resolution reconstruction models are described, the advantages and disadvantages of different algorithms and the applicable application scenarios are analyzed and compared, and the different types of super-resolution algorithms are discussed. The method of mutual fusion and image and video quality evaluation, and a brief introduction to commonly used data sets. Finally, the potential problems faced by the current image super-resolution reconstruction technology are discussed, and a new outlook for the future development direction is made.