{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/filmy-cloud-removal-on-satellite-imagery-with","title":"Filmy Cloud Removal on Satellite Imagery with Multispectral Conditional Generative Adversarial Nets","arxiv_id":"1710.04835","date":"2017-10-13","proceeding":null,"authors":["Kenji Enomoto","Ken Sakurada","Weimin WANG","Hiroshi Fukui","Masashi Matsuoka","Ryosuke Nakamura","Nobuo Kawaguchi"],"abstract":"In this paper, we propose a method for cloud removal from visible light RGB\nsatellite images by extending the conditional Generative Adversarial Networks\n(cGANs) from RGB images to multispectral images. Satellite images have been\nwidely utilized for various purposes, such as natural environment monitoring\n(pollution, forest or rivers), transportation improvement and prompt emergency\nresponse to disasters. However, the obscurity caused by clouds makes it\nunstable to monitor the situation on the ground with the visible light camera.\nImages captured by a longer wavelength are introduced to reduce the effects of\nclouds. Synthetic Aperture Radar (SAR) is such an example that improves\nvisibility even the clouds exist. On the other hand, the spatial resolution\ndecreases as the wavelength increases. Furthermore, the images captured by long\nwavelengths differs considerably from those captured by visible light in terms\nof their appearance. Therefore, we propose a network that can remove clouds and\ngenerate visible light images from the multispectral images taken as inputs.\nThis is achieved by extending the input channels of cGANs to be compatible with\nmultispectral images. The networks are trained to output images that are close\nto the ground truth using the images synthesized with clouds over the ground\ntruth as inputs. In the available dataset, the proportion of images of the\nforest or the sea is very high, which will introduce bias in the training\ndataset if uniformly sampled from the original dataset. Thus, we utilize the\nt-Distributed Stochastic Neighbor Embedding (t-SNE) to improve the problem of\nbias in the training dataset. Finally, we confirm the feasibility of the\nproposed network on the dataset of four bands images, which include three\nvisible light bands and one near-infrared (NIR) band.","url_abs":"http://arxiv.org/abs/1710.04835v1","url_pdf":"http://arxiv.org/pdf/1710.04835v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[],"tasks":[{"task_slug":"cloud-removal","task_name":"Cloud Removal"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/cloud-removal-on-sen12ms-cr","task":"Cloud Removal","dataset":"SEN12MS-CR","model":"McGAN","rank_in_archive_order":10,"of":10,"metrics":{"MAE":"0.048","PSNR":"25.14","SAM":"15.676","SSIM":"0.744"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1710.04835","atlas_url":"https://app.syntology.ai/?focus=1710.04835","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}