Single Image Super-Resolution with Application to Remote-Sensing Image

Abstract

To improve the resolution of satellite images, many researchers are committed to machine learning and neural network-based SR methods. SR has multiple residual network frameworks in deep learning that have improved performance and can extend thousands of layers in the system. However, each layer improves accuracy by doubling the number of layers, although training thousands of layers are too expensive, the process is slow, and there are functional recovery issues. To address these issues, we propose a super-resolution wide remote sensing residual network (WRSR), in which we increase the width and reduce the depth of the residual network, due to decreasing the depth of the network our model reduced memory costs. To enhance the resolution of the single image we showed that our method improves training loss performance by performing the weight normalization instead of augmentation technology. The results of the experiment show that the method performs well in terms of quantitative indicators (PSNR) and (SSIM).