Image Enhancement and Improvement Algorithm Based on Esrgan Singal Frame Remote Sensing Image

Traditional spline interpolation algorithm for reconstruction of visual effects are not good, based on Super Resolution against Network (Super - Resolution Generative Adversarial Network, SRGAN) edge character to deal with such problems as imperfect, using a generated based on the enhanced Super Resolution against Network to improve the Resolution of ordinary optical remote sensing images, the first Network generated by high Resolution data sets to train (G network), then lower Resolution test data model test, The test results and real results are put into the discrimination network (D network) to get the adversarial loss, and then the generated network is modified according to the adversarial loss. The superiority of the network is improved by introducing dense residuals to SRGAN, modifying the judgment object of the discriminator to be relatively real, and using the eigenvalue before activation to improve the perceived loss. The desert, farmland, forest and mountain data were tested on AID data set, and the algorithm in this paper could obtain the recomposition of the real image more closely. Compared with SRResNet and SRGAN algorithms, PSNR improved by about 4.0db and SSIM improved by about 0.14. This method improves the feature comprehensiveness by increasing the network fineness degree, and USES the modified perception loss to get the brightness closer to the real image, which is beneficial to improve the quality of single frame remote sensing image.