{"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/an-artificial-agent-for-robust-image","title":"An Artificial Agent for Robust Image Registration","arxiv_id":"1611.10336","date":"2016-11-30","proceeding":null,"authors":["Rui Liao","Shun Miao","Pierre de Tournemire","Sasa Grbic","Ali Kamen","Tommaso Mansi","Dorin Comaniciu"],"abstract":"3-D image registration, which involves aligning two or more images, is a\ncritical step in a variety of medical applications from diagnosis to therapy.\nImage registration is commonly performed by optimizing an image matching metric\nas a cost function. However, this task is challenging due to the non-convex\nnature of the matching metric over the plausible registration parameter space\nand insufficient approaches for a robust optimization. As a result, current\napproaches are often customized to a specific problem and sensitive to image\nquality and artifacts. In this paper, we propose a completely different\napproach to image registration, inspired by how experts perform the task. We\nfirst cast the image registration problem as a \"strategy learning\" process,\nwhere the goal is to find the best sequence of motion actions (e.g. up, down,\netc.) that yields image alignment. Within this approach, an artificial agent is\nlearned, modeled using deep convolutional neural networks, with 3D raw image\ndata as the input, and the next optimal action as the output. To cope with the\ndimensionality of the problem, we propose a greedy supervised approach for an\nend-to-end training, coupled with attention-driven hierarchical strategy. The\nresulting registration approach inherently encodes both a data-driven matching\nmetric and an optimal registration strategy (policy). We demonstrate, on two\n3-D/3-D medical image registration examples with drastically different nature\nof challenges, that the artificial agent outperforms several state-of-art\nregistration methods by a large margin in terms of both accuracy and\nrobustness.","url_abs":"http://arxiv.org/abs/1611.10336v1","url_pdf":"http://arxiv.org/pdf/1611.10336v1.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":"an-artificial-agent-for-robust-image","repo_url":"https://github.com/nicolas1805961/GE_Reinforcement_Learning_Image_Registration","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"image-registration","task_name":"Image Registration"},{"task_slug":"medical-image-registration","task_name":"Medical Image Registration"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1611.10336","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}