Laplacian Pyramid Adaptive Residual Network Satellite Image Super-Resolution

Authors

  • Zikang Wei , Yunqing Liu , Xuan Li

Abstract

Aiming at the application of super-resolution reconstruction of Tiangong-1 satellite image, a deep structure model of Laplacian pyramid adaptive residual network is proposed. In the research of this paper, the popular residual module is improved to make full use of the feature information with all convolutional layer in the residual block to improve the image quality; propose the Laplacian pyramid structure, deepen the network structure and strengthen the image features extraction ability, thereby extracting more image features; propose feature adaptive selection structure, through the feature adaptive selection structure layer to adaptively optimize the image features, enhance the quality of the reconstructed image. The experiment of Tiangong-1 images shows that the Laplacian pyramid adaptive residual network proposed in the study has high PSNR and SSIM values in the application of satellite image super-resolution reconstruction, and the band shift is smaller, which has high application value and deeper research significance.

Published

2020-04-30

Issue

Section

Articles