{"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-converged-propagations-with-deep","title":"Learning Converged Propagations with Deep Prior Ensemble for Image Enhancement","arxiv_id":"1810.04012","date":"2018-10-09","proceeding":null,"authors":["Risheng Liu","Long Ma","Yiyang Wang","Lei Zhang"],"abstract":"Enhancing visual qualities of images plays very important roles in various\nvision and learning applications. In the past few years, both knowledge-driven\nmaximum a posterior (MAP) with prior modelings and fully data-dependent\nconvolutional neural network (CNN) techniques have been investigated to address\nspecific enhancement tasks. In this paper, by exploiting the advantages of\nthese two types of mechanisms within a complementary propagation perspective,\nwe propose a unified framework, named deep prior ensemble (DPE), for solving\nvarious image enhancement tasks. Specifically, we first establish the basic\npropagation scheme based on the fundamental image modeling cues and then\nintroduce residual CNNs to help predicting the propagation direction at each\nstage. By designing prior projections to perform feedback control, we\ntheoretically prove that even with experience-inspired CNNs, DPE is definitely\nconverged and the output will always satisfy our fundamental task constraints.\nThe main advantage against conventional optimization-based MAP approaches is\nthat our descent directions are learned from collected training data, thus are\nmuch more robust to unwanted local minimums. While, compared with existing CNN\ntype networks, which are often designed in heuristic manners without\ntheoretical guarantees, DPE is able to gain advantages from rich task cues\ninvestigated on the bases of domain knowledges. Therefore, DPE actually\nprovides a generic ensemble methodology to integrate both knowledge and\ndata-based cues for different image enhancement tasks. More importantly, our\ntheoretical investigations verify that the feedforward propagations of DPE are\nproperly controlled toward our desired solution. Experimental results\ndemonstrate that the proposed DPE outperforms state-of-the-arts on a variety of\nimage enhancement tasks in terms of both quantitative measure and visual\nperception quality.","url_abs":"http://arxiv.org/abs/1810.04012v1","url_pdf":"http://arxiv.org/pdf/1810.04012v1.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-converged-propagations-with-deep","repo_url":"https://github.com/dlut-dimt/DPE-Deep-Prior-Ensemble","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"image-enhancement","task_name":"Image Enhancement"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}