{"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/the-conditional-lucas-kanade-algorithm","title":"The Conditional Lucas & Kanade Algorithm","arxiv_id":"1603.08597","date":"2016-03-29","proceeding":null,"authors":["Chen-Hsuan Lin","Rui Zhu","Simon Lucey"],"abstract":"The Lucas & Kanade (LK) algorithm is the method of choice for efficient dense\nimage and object alignment. The approach is efficient as it attempts to model\nthe connection between appearance and geometric displacement through a linear\nrelationship that assumes independence across pixel coordinates. A drawback of\nthe approach, however, is its generative nature. Specifically, its performance\nis tightly coupled with how well the linear model can synthesize appearance\nfrom geometric displacement, even though the alignment task itself is\nassociated with the inverse problem. In this paper, we present a new approach,\nreferred to as the Conditional LK algorithm, which: (i) directly learns linear\nmodels that predict geometric displacement as a function of appearance, and\n(ii) employs a novel strategy for ensuring that the generative pixel\nindependence assumption can still be taken advantage of. We demonstrate that\nour approach exhibits superior performance to classical generative forms of the\nLK algorithm. Furthermore, we demonstrate its comparable performance to\nstate-of-the-art methods such as the Supervised Descent Method with\nsubstantially less training examples, as well as the unique ability to \"swap\"\ngeometric warp functions without having to retrain from scratch. Finally, from\na theoretical perspective, our approach hints at possible redundancies that\nexist in current state-of-the-art methods for alignment that could be leveraged\nin vision systems of the future.","url_abs":"http://arxiv.org/abs/1603.08597v1","url_pdf":"http://arxiv.org/pdf/1603.08597v1.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":"the-conditional-lucas-kanade-algorithm","repo_url":"https://github.com/chenhsuanlin/conditional-LK","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}