Single Image Super-Resolution Based on Enhanced Residual Neural Network

Vision is an important way for humans to perceive the world around them. It is an urgent need to improve image quality by using image super resolution. Residual network plays an important role in image super resolution because of its powerful feature extraction ability. Only stacking residual network is not enough to extract a large amount of effective information, and at the same time, more training memory is needed. Therefore, this paper designs a single feature extraction module with better performance to enhance the feature extraction ability of the network. In this paper, the Residual Network is mainly improved, and a Single Image Super-Resolution Based on Enhanced Residual Neural Network (ERNN) is proposed based on improved residual module. Compared with other classical lightweight algorithms, the lightweight network with the new residual module as the basic unit has higher reconstruction ability, and can achieve better image super-resolution reconstruction with a small number of parameters. Moreover, a large number of experiments have verified that the proposed ERNN can achieve the trade-off between the number of parameters and performance.