Super-resolution reconstruction of remote sensing images based on Local Discriminative Learning and Pyramid Squeeze Attention

Due to the physical factors such as imaging distance and imaging period of remote sensing system, the resolution of remote sensing images is usually low. The existing image super-resolution reconstruction methods cannot effectively use multi-scale information, and the reconstructed image is prone to visual artifacts, loss of high-frequency information and other problems. To solve this problem, a super-resolution reconstruction network of remote sensing images is designed based on generative adversarial network, which combines local discriminative learning and pyramid squeezing attention. The network is composed of local discriminative learning (LDL), pyramid squeeze attention module and residuals dense connection. The generated network extracts the detailed features of remote sensing images through residual dense connection blocks and adds a pyramid squeeze attention mechanism (PSA) to effectively extract multi-scale spatial information. Real details are generated through local discriminative learning, which effectively suppressed the generation of visual artifacts. The experimental results demonstrate that the proposed method can effectively improve the resolution of remote sensing images and suppress visual artifacts.