{"title":"Super-resolution reconstruction of remote sensing images based on SRGAN","authors":"Meng Yu, Hongjuan Wang, Chang Liu, Deping Lin","doi":"10.1117/12.2653847","DOIUrl":null,"url":null,"abstract":"In the field of remote sensing images, due to the limitations of hardware equipment, image transmission, natural environment and other reasons, the resolution of the obtained remote sensing images cannot reach the desired resolution. The emergence of image super-resolution reconstruction technology can improve the resolution of remote sensing images without increasing the high cost. Image super-resolution reconstruction refers to the fact that low-resolution images can obtain high-resolution images through certain algorithmic techniques. With the rapid development of deep learning ideas, researchers have applied it to the field of image super-resolution reconstruction and achieved good results. Image super-resolution reconstruction also shifts from traditional reconstruction methods to deep learning-based methods. The emergence of the idea of Generative Adversarial Networks has further advanced the field of image super-resolution reconstruction. By using the idea of Generative Adversarial Network (GAN), researchers can obtain more realistic high-resolution images. This paper mainly uses the SRGAN model, the image dataset DIV2K for super-resolution reconstruction, and uses a dense residual structure in the generator network to obtain more image information, so that the effect of image reconstruction is more realistic. Through the experimental verification on the SIRI-WHU remote sensing test data set, the two evaluation indicators of peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) are compared, and the effect is improved. Better generation results can also be observed through subjective human vision.","PeriodicalId":253792,"journal":{"name":"Conference on Optics and Communication Technology","volume":"28 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Conference on Optics and Communication Technology","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1117/12.2653847","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
In the field of remote sensing images, due to the limitations of hardware equipment, image transmission, natural environment and other reasons, the resolution of the obtained remote sensing images cannot reach the desired resolution. The emergence of image super-resolution reconstruction technology can improve the resolution of remote sensing images without increasing the high cost. Image super-resolution reconstruction refers to the fact that low-resolution images can obtain high-resolution images through certain algorithmic techniques. With the rapid development of deep learning ideas, researchers have applied it to the field of image super-resolution reconstruction and achieved good results. Image super-resolution reconstruction also shifts from traditional reconstruction methods to deep learning-based methods. The emergence of the idea of Generative Adversarial Networks has further advanced the field of image super-resolution reconstruction. By using the idea of Generative Adversarial Network (GAN), researchers can obtain more realistic high-resolution images. This paper mainly uses the SRGAN model, the image dataset DIV2K for super-resolution reconstruction, and uses a dense residual structure in the generator network to obtain more image information, so that the effect of image reconstruction is more realistic. Through the experimental verification on the SIRI-WHU remote sensing test data set, the two evaluation indicators of peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) are compared, and the effect is improved. Better generation results can also be observed through subjective human vision.