{"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/deep-bilateral-learning-for-real-time-image","title":"Deep Bilateral Learning for Real-Time Image Enhancement","arxiv_id":"1707.02880","date":"2017-07-10","proceeding":null,"authors":["Michaël Gharbi","Jiawen Chen","Jonathan T. Barron","Samuel W. Hasinoff","Frédo Durand"],"abstract":"Performance is a critical challenge in mobile image processing. Given a\nreference imaging pipeline, or even human-adjusted pairs of images, we seek to\nreproduce the enhancements and enable real-time evaluation. For this, we\nintroduce a new neural network architecture inspired by bilateral grid\nprocessing and local affine color transforms. Using pairs of input/output\nimages, we train a convolutional neural network to predict the coefficients of\na locally-affine model in bilateral space. Our architecture learns to make\nlocal, global, and content-dependent decisions to approximate the desired image\ntransformation. At runtime, the neural network consumes a low-resolution\nversion of the input image, produces a set of affine transformations in\nbilateral space, upsamples those transformations in an edge-preserving fashion\nusing a new slicing node, and then applies those upsampled transformations to\nthe full-resolution image. Our algorithm processes high-resolution images on a\nsmartphone in milliseconds, provides a real-time viewfinder at 1080p\nresolution, and matches the quality of state-of-the-art approximation\ntechniques on a large class of image operators. Unlike previous work, our model\nis trained off-line from data and therefore does not require access to the\noriginal operator at runtime. This allows our model to learn complex,\nscene-dependent transformations for which no reference implementation is\navailable, such as the photographic edits of a human retoucher.","url_abs":"http://arxiv.org/abs/1707.02880v2","url_pdf":"http://arxiv.org/pdf/1707.02880v2.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":"deep-bilateral-learning-for-real-time-image","repo_url":"https://github.com/google/hdrnet","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"jax","reach":{"status":"unanswered"}},{"paper_slug":"deep-bilateral-learning-for-real-time-image","repo_url":"https://github.com/creotiv/hdrnet-pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"image-enhancement","task_name":"Image Enhancement"},{"task_slug":"image-retouching","task_name":"Image Retouching"}],"methods":[{"method_slug":"bilateral-grid","method_name":"Bilateral Grid"}],"datasets_introduced":[],"methods_introduced":[{"slug":"bilateral-grid","name":"Bilateral Grid","full_name":"Bilateral Grid"}],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1707.02880","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}