{"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/generating-training-data-for-denoising-real","title":"Generating Training Data for Denoising Real RGB Images via Camera Pipeline Simulation","arxiv_id":"1904.08825","date":"2019-04-18","proceeding":null,"authors":["Ronnachai Jaroensri","Camille Biscarrat","Miika Aittala","Frédo Durand"],"abstract":"Image reconstruction techniques such as denoising often need to be applied to\nthe RGB output of cameras and cellphones. Unfortunately, the commonly used\nadditive white noise (AWGN) models do not accurately reproduce the noise and\nthe degradation encountered on these inputs. This is particularly important for\nlearning-based techniques, because the mismatch between training and real world\ndata will hurt their generalization. This paper aims to accurately simulate the\ndegradation and noise transformation performed by camera pipelines. This allows\nus to generate realistic degradation in RGB images that can be used to train\nmachine learning models. We use our simulation to study the importance of noise\nmodeling for learning-based denoising. Our study shows that a realistic noise\nmodel is required for learning to denoise real JPEG images. A neural network\ntrained on realistic noise outperforms the one trained with AWGN by 3 dB. An\nablation study of our pipeline shows that simulating denoising and demosaicking\nis important to this improvement and that realistic demosaicking algorithms,\nwhich have been rarely considered, is needed. We believe this simulation will\nalso be useful for other image reconstruction tasks, and we will distribute our\ncode publicly.","url_abs":"http://arxiv.org/abs/1904.08825v1","url_pdf":"http://arxiv.org/pdf/1904.08825v1.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":"generating-training-data-for-denoising-real","repo_url":"https://github.com/12dmodel/camera_sim","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"demosaicking","task_name":"Demosaicking"},{"task_slug":"denoising","task_name":"Denoising"},{"task_slug":"image-reconstruction","task_name":"Image Reconstruction"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}