Deep Residual Network for Image Super-Resolution Reconstruction
Bozhen Zhang, Cheng Gao · 2022 12th International Conference on CYBER Technology in Automation, Control, and Intelligent Systems (CYBER) · 2022
In view of the fact that traditional neural networks mostly only use the spatial domain information of images over a period of single image super-resolution reconstruction, the generated images are prone to lose important details. We propose a method based on deep residual networks. The sub-image is used as the input of the deep residual network, and then the residual network is improved by combining both local and global residual learning, which reduces the phenomenon of gradient disappearance and improves the efficiency of information transfer. The multi-scale convolution kernel is used to extract rich feature texture information for obtaining the reconstructed image. The experimental results indicate that the reconstruction effect of this method achieves better subjective visual experience and objective evaluation indicators than the contrasting algorithms.