{"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/image-restoration-using-very-deep","title":"Image Restoration Using Very Deep Convolutional Encoder-Decoder Networks with Symmetric Skip Connections","arxiv_id":"1603.09056","date":"2016-03-30","proceeding":"NeurIPS 2016 12","authors":["Xiao-Jiao Mao","Chunhua Shen","Yu-Bin Yang"],"abstract":"In this paper, we propose a very deep fully convolutional encoding-decoding\nframework for image restoration such as denoising and super-resolution. The\nnetwork is composed of multiple layers of convolution and de-convolution\noperators, learning end-to-end mappings from corrupted images to the original\nones. The convolutional layers act as the feature extractor, which capture the\nabstraction of image contents while eliminating noises/corruptions.\nDe-convolutional layers are then used to recover the image details. We propose\nto symmetrically link convolutional and de-convolutional layers with skip-layer\nconnections, with which the training converges much faster and attains a\nhigher-quality local optimum. First, The skip connections allow the signal to\nbe back-propagated to bottom layers directly, and thus tackles the problem of\ngradient vanishing, making training deep networks easier and achieving\nrestoration performance gains consequently. Second, these skip connections pass\nimage details from convolutional layers to de-convolutional layers, which is\nbeneficial in recovering the original image. Significantly, with the large\ncapacity, we can handle different levels of noises using a single model.\nExperimental results show that our network achieves better performance than all\npreviously reported state-of-the-art methods.","url_abs":"http://arxiv.org/abs/1603.09056v2","url_pdf":"http://arxiv.org/pdf/1603.09056v2.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":[{"paper_slug":"image-restoration-using-very-deep","repo_url":"https://github.com/andrewhproppe/phaseretrievalnns","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"image-restoration-using-very-deep","repo_url":"https://github.com/lyqcom/rednet30","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"mindspore","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"image-restoration-using-very-deep","repo_url":"https://github.com/yjn870/REDNet-pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"decoder","task_name":"Decoder"},{"task_slug":"denoising","task_name":"Denoising"},{"task_slug":"image-restoration","task_name":"Image Restoration"},{"task_slug":"image-super-resolution","task_name":"Image Super-Resolution"},{"task_slug":"super-resolution","task_name":"Super-Resolution"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/image-super-resolution-on-bsd100-4x-upscaling","task":"Image Super-Resolution","dataset":"BSD100 - 4x upscaling","model":"RED30","rank_in_archive_order":43,"of":71,"metrics":{"PSNR":"27.40","SSIM":"0.7290"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1603.09056","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}