{"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/learning-jpeg-compression-artifacts-for-image","title":"Learning JPEG Compression Artifacts for Image Manipulation Detection and Localization","arxiv_id":"2108.12947","date":"2021-08-30","proceeding":null,"authors":["Myung-Joon Kwon","Seung-Hun Nam","In-Jae Yu","Heung-Kyu Lee","Changick Kim"],"abstract":"Detecting and localizing image manipulation are necessary to counter malicious use of image editing techniques. Accordingly, it is essential to distinguish between authentic and tampered regions by analyzing intrinsic statistics in an image. We focus on JPEG compression artifacts left during image acquisition and editing. We propose a convolutional neural network (CNN) that uses discrete cosine transform (DCT) coefficients, where compression artifacts remain, to localize image manipulation. Standard CNNs cannot learn the distribution of DCT coefficients because the convolution throws away the spatial coordinates, which are essential for DCT coefficients. We illustrate how to design and train a neural network that can learn the distribution of DCT coefficients. Furthermore, we introduce Compression Artifact Tracing Network (CAT-Net) that jointly uses image acquisition artifacts and compression artifacts. It significantly outperforms traditional and deep neural network-based methods in detecting and localizing tampered regions.","url_abs":"https://arxiv.org/abs/2108.12947v2","url_pdf":"https://arxiv.org/pdf/2108.12947v2.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":"learning-jpeg-compression-artifacts-for-image","repo_url":"https://github.com/mjkwon2021/cat-net","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"image-manipulation","task_name":"Image Manipulation"},{"task_slug":"image-manipulation-detection","task_name":"Image Manipulation Detection"},{"task_slug":"image-manipulation-localization","task_name":"Image Manipulation Localization"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"discrete-cosine-transform","method_name":"Discrete Cosine Transform"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/image-manipulation-detection-on-coverage","task":"Image Manipulation Detection","dataset":"COVERAGE","model":"CAT-Net v2","rank_in_archive_order":4,"of":8,"metrics":{"AUC":".680","Balanced Accuracy":".635"},"uses_additional_data":false},{"leaderboard":"/sota/image-manipulation-detection-on-casia-v1","task":"Image Manipulation Detection","dataset":"Casia V1+","model":"CAT-Net v2","rank_in_archive_order":3,"of":9,"metrics":{"AUC":".942","Balanced Accuracy":".838"},"uses_additional_data":false},{"leaderboard":"/sota/image-manipulation-detection-on-cocoglide","task":"Image Manipulation Detection","dataset":"CocoGlide","model":"CAT-Net v2","rank_in_archive_order":4,"of":8,"metrics":{"AUC":".667","Balanced Accuracy":".580"},"uses_additional_data":false},{"leaderboard":"/sota/image-manipulation-detection-on-columbia","task":"Image Manipulation Detection","dataset":"Columbia","model":"CAT-Net v2","rank_in_archive_order":4,"of":8,"metrics":{"AUC":".977","Balanced Accuracy":".803"},"uses_additional_data":false},{"leaderboard":"/sota/image-manipulation-detection-on-dso-1","task":"Image Manipulation Detection","dataset":"DSO-1","model":"CAT-Net v2","rank_in_archive_order":4,"of":9,"metrics":{"AUC":".747","Balanced Accuracy":".525"},"uses_additional_data":false},{"leaderboard":"/sota/image-manipulation-localization-on-casiav1","task":"Image Manipulation Localization","dataset":"CASIAv1(Protoclo-CAT)","model":"CAT-Net","rank_in_archive_order":3,"of":8,"metrics":{"Pixel Binary F1":"0.808"},"uses_additional_data":false},{"leaderboard":"/sota/image-manipulation-localization-on-coverage","task":"Image Manipulation Localization","dataset":"COVERAGE","model":"CAT-Net v2","rank_in_archive_order":9,"of":11,"metrics":{"Average Pixel F1(Fixed threshold)":".381"},"uses_additional_data":false},{"leaderboard":"/sota/image-manipulation-localization-on-coverage-1","task":"Image Manipulation Localization","dataset":"COVERAGE(Protocol-CAT)","model":"CAT-Net","rank_in_archive_order":5,"of":8,"metrics":{"Pixel Binary F1":"0.427"},"uses_additional_data":false},{"leaderboard":"/sota/image-manipulation-localization-on-casia-v1","task":"Image Manipulation Localization","dataset":"Casia V1+","model":"CAT-Net v2","rank_in_archive_order":6,"of":11,"metrics":{"Average Pixel F1(Fixed threshold)":".752"},"uses_additional_data":false},{"leaderboard":"/sota/image-manipulation-localization-on-cocoglide","task":"Image Manipulation Localization","dataset":"CocoGlide","model":"CAT-Net v2","rank_in_archive_order":10,"of":11,"metrics":{"Average Pixel F1(Fixed threshold)":".434"},"uses_additional_data":false},{"leaderboard":"/sota/image-manipulation-localization-on-columbia","task":"Image Manipulation Localization","dataset":"Columbia","model":"CAT-Net v2","rank_in_archive_order":6,"of":11,"metrics":{"Average Pixel F1(Fixed threshold)":".859"},"uses_additional_data":false},{"leaderboard":"/sota/image-manipulation-localization-on-columbia-1","task":"Image Manipulation Localization","dataset":"Columbia(Protocol-CAT)","model":"CAT-Net","rank_in_archive_order":3,"of":8,"metrics":{"Pixel Binary F1":"0.915"},"uses_additional_data":false},{"leaderboard":"/sota/image-manipulation-localization-on-dso-1","task":"Image Manipulation Localization","dataset":"DSO-1","model":"CAT-Net v2","rank_in_archive_order":7,"of":11,"metrics":{"Average Pixel F1(Fixed threshold)":".584"},"uses_additional_data":false},{"leaderboard":"/sota/image-manipulation-localization-on-nist16","task":"Image Manipulation Localization","dataset":"NIST16(Protocol-CAT)","model":"CAT-Net","rank_in_archive_order":6,"of":8,"metrics":{"Pixel Binary F1":"0.252"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2108.12947","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}