{"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/trainable-nonlinear-reaction-diffusion-a","title":"Trainable Nonlinear Reaction Diffusion: A Flexible Framework for Fast and Effective Image Restoration","arxiv_id":"1508.02848","date":"2015-08-12","proceeding":null,"authors":["Yunjin Chen","Thomas Pock"],"abstract":"Image restoration is a long-standing problem in low-level computer vision\nwith many interesting applications. We describe a flexible learning framework\nbased on the concept of nonlinear reaction diffusion models for various image\nrestoration problems. By embodying recent improvements in nonlinear diffusion\nmodels, we propose a dynamic nonlinear reaction diffusion model with\ntime-dependent parameters (\\ie, linear filters and influence functions). In\ncontrast to previous nonlinear diffusion models, all the parameters, including\nthe filters and the influence functions, are simultaneously learned from\ntraining data through a loss based approach. We call this approach TNRD --\n\\textit{Trainable Nonlinear Reaction Diffusion}. The TNRD approach is\napplicable for a variety of image restoration tasks by incorporating\nappropriate reaction force. We demonstrate its capabilities with three\nrepresentative applications, Gaussian image denoising, single image super\nresolution and JPEG deblocking. Experiments show that our trained nonlinear\ndiffusion models largely benefit from the training of the parameters and\nfinally lead to the best reported performance on common test datasets for the\ntested applications. Our trained models preserve the structural simplicity of\ndiffusion models and take only a small number of diffusion steps, thus are\nhighly efficient. Moreover, they are also well-suited for parallel computation\non GPUs, which makes the inference procedure extremely fast.","url_abs":"http://arxiv.org/abs/1508.02848v2","url_pdf":"http://arxiv.org/pdf/1508.02848v2.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":"denoising","task_name":"Denoising"},{"task_slug":"image-denoising","task_name":"Image 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":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/color-image-denoising-on-darmstadt-noise","task":"Color Image Denoising","dataset":"Darmstadt Noise Dataset","model":"TNRD","rank_in_archive_order":6,"of":6,"metrics":{"PSNR (sRGB)":"33.65","SSIM (sRGB)":"0.8306"},"uses_additional_data":false},{"leaderboard":"/sota/denoising-on-darmstadt-noise-dataset","task":"Denoising","dataset":"Darmstadt Noise Dataset","model":"TNRD","rank_in_archive_order":8,"of":10,"metrics":{"PSNR":"33.65"},"uses_additional_data":false},{"leaderboard":"/sota/grayscale-image-denoising-on-bsd68-sigma15","task":"Grayscale Image Denoising","dataset":"BSD68 sigma15","model":"TNRD","rank_in_archive_order":13,"of":16,"metrics":{"PSNR":"31.42"},"uses_additional_data":false},{"leaderboard":"/sota/grayscale-image-denoising-on-bsd68-sigma25","task":"Grayscale Image Denoising","dataset":"BSD68 sigma25","model":"TNRD","rank_in_archive_order":15,"of":16,"metrics":{"PSNR":"28.92"},"uses_additional_data":false},{"leaderboard":"/sota/grayscale-image-denoising-on-urban100-sigma15","task":"Grayscale Image Denoising","dataset":"Urban100 sigma15","model":"TNRD","rank_in_archive_order":7,"of":7,"metrics":{"PSNR":"31.98"},"uses_additional_data":false},{"leaderboard":"/sota/image-super-resolution-on-set14-4x-upscaling","task":"Image Super-Resolution","dataset":"Set14 - 4x upscaling","model":"TNRD","rank_in_archive_order":87,"of":104,"metrics":{"PSNR":"27.68"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1508.02848","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}