[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86281-en":3,"doc-seo-86281-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},86281,1099514068365,"Aurelia","https://ap-avatar.wpscdn.com/avatar/10000253d8d9f28188e?_k=1776742907772140068",8,"Research & Report","Backbone-Agnostic Perturbation-Induced Uncertainty Learning for End-to-End Real-World Image Dehazing","Real-world paired image dehazing remains difficult due to spatially non-uniform haze, illumination dependence, and physical ambiguity even when haze-free references exist. Existing end-to-end restoration networks typically use deterministic hazy-to-clean mappings, leaving uncertainty in features, priors, and cross-domain negatives underexplored. This work introduces BPUL, a plug-and-play framework with LPUM for learnable perturbation-based uncertainty, PURM enforcing degradation consistency via transmission and atmospheric-light priors, and D3CL regularizing dual spaces using real and synthetic negatives. Experiments on five paired benchmarks show improved restoration across multiple backbones with minimal inference overhead.","Backbone-Agnostic Perturbation-Induced Uncertainty Learning for End-to-End  \nReal-World Image Dehazing  \nBingcai Wei  \nSchool of Computer Science  \nWuhan University  \nWuhan, Hubei, China  \n[weibc97@whu.edu.cn](weibc97@whu.edu.cn)  \narXiv :2607 . 11623v1 [ cs .CV] 13 Jul 2026  \nAbstract  \nReal-world paired image dehazing remains challenging because haze degradation is spatially non-uniform, illuminationdependent, and physically ambiguous even when haze-free references are available. Existing end-to-end restoration networks usually formulate dehazing as a deterministic mapping from a hazy observation to a clean target, leaving the uncertainty hidden in degraded features, haze priors, and crossdomain negative samples insufficiently explored. In this paper, we propose Backbone-Agnostic Perturbation-Induced Uncertainty Learning (BPUL), a plug-and-play uncertainty learning framework for end-to-end real-world image dehazing. BPUL first introduces a Learnable Perturbation-induced Uncertainty Modulator (LPUM) that estimates channel-wise and spatialwise feature sensitivity through reparameterized stochastic perturbations. It then develops a Prior-informed Uncertaintyguided Reconstruction Module (PURM), which exploits transmission and atmospheric-light priors to reconstruct the hazy observation from the restored result and enforce degradation consistency. Furthermore, we propose a Dual-space Domaindiversified Distribution-aware Contrastive Loss (D3CL) to regularize both clean restoration and hazy reconstruction spaces with real-world and synthetic negatives. Experiments on five real-world paired benchmarks show that BPUL consistently improves multiple representative backbones. Since only LPUM is retained during inference while PURM and D3CLare used as training-time constraints, BPUL brings substantial restoration gains with only marginal additional inference overhead.  \nIntroduction  \nImage dehazing aims to recover a visually faithful haze-free image from a hazy observation and remains a fundamental ill-posed low-level vision problem. The classical atmospheric scattering model describes a hazy image as  \nI (x) = J (x)t (x) + A(1 − t(x)), (1)  \nwhere I (x) and J (x) are the hazy observation and hazefree radiance, t (x) is the transmission map, and A is the global atmospheric light. Although this formulation motivates interpretable priors such as the dark channel prior (He, Sun, and Tang 2011), real haze rarely follows a deterministic pattern. Spatially varying density, illumination, depth ambiguity, camera response, and prior-estimation errors make real-world dehazing substantially more uncertain than standard supervised restoration.  \nReal-world paired datasets, including I-HAZE (Ancutiet al. 2018a), O-HAZE (Ancuti et al. 2018b), Dense-Haze (Ancuti et al. 2019), NH-HAZE (Ancuti, Ancuti, and Timofte 2020), and LMHaze (Zhang et al. 2024), enable quantitative evaluation with aligned hazy and haze-free pairs, yet also expose the ambiguity that paired supervision cannot remove. Dense haze suppresses texture and contrast, nonhomogeneous haze varies spatially, and multi-intensity haze creates strong cross-scene shifts. Consequently, pixel-wise reconstruction alone can yield over-smoothed structures or color deviations despite visually plausible outputs.  \nImage restoration has progressed through stronger CNN and Transformer backbones (Cho et al. 2021; Chen et al. 2022a; Zamir et al. 2022; Wang et al. 2022; Cui, Ren, and Knoll 2024; Wu et al. 2026b; Cui et al. 2026), while dehazing methods increasingly exploit multi-scale feedback, color guidance, contrastive regularization, and uncertainty-aware processing (Dong et al. 2020; Wu et al. 2021; Zheng et al. 2023; Hong et al. 2022; Fang et al. 2025; Liu et al. 2025) . Nevertheless, most end-to-end systems learn a deterministic mapping ˆJ = fθ (I) dominated by restoration error. They do not explicitly identify degradation-sensitive channels and locations, test whether the restored image explains the ","cbCaijPA5L95ftEP","https://ap.wps.com/l/cbCaijPA5L95ftEP","pdf",2699890,2,1,9,"English","en",105,"# Abstract\n# Introduction","[{\"question\":\"Why is real-world paired image dehazing still challenging even with haze-free references?\",\"answer\":\"Haze degradation is spatially non-uniform, illumination-dependent, and physically ambiguous, and real haze rarely follows a deterministic pattern. These factors create uncertainty that paired supervision cannot fully eliminate.\"},{\"question\":\"What is BPUL and how does it model uncertainty for dehazing?\",\"answer\":\"BPUL introduces LPUM to estimate channel-wise and spatial-wise feature sensitivity using reparameterized stochastic perturbations. Uncertainty is converted into feature modulation signals to adapt restoration to degradation-sensitive regions and channels.\"},{\"question\":\"How do PURM and D3CL contribute during training, and what is kept for inference?\",\"answer\":\"PURM reconstructs the hazy observation from the restored result using transmission and atmospheric-light priors to enforce degradation consistency. D3CL regularizes clean restoration and hazy reconstruction spaces with real and synthetic negatives. During inference, only LPUM is retained, while PURM and D3CL are training-time constraints.\"}]",1784210017,23,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"backbone-agnostic-perturbation-induced-uncertainty-learning-for-end-to-end-real-world-image-dehazing","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,47,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":20},"https://docshare.wps.com/document/","Document",{"item":48,"name":12,"@type":43,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/backbone-agnostic-perturbation-induced-uncertainty-learning-for-end-to-end-real-world-image-dehazing/86281/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-27","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why is real-world paired image dehazing still challenging even with haze-free references?","Question",{"text":75,"@type":76},"Haze degradation is spatially non-uniform, illumination-dependent, and physically ambiguous, and real haze rarely follows a deterministic pattern. These factors create uncertainty that paired supervision cannot fully eliminate.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is BPUL and how does it model uncertainty for dehazing?",{"text":80,"@type":76},"BPUL introduces LPUM to estimate channel-wise and spatial-wise feature sensitivity using reparameterized stochastic perturbations. Uncertainty is converted into feature modulation signals to adapt restoration to degradation-sensitive regions and channels.",{"name":82,"@type":73,"acceptedAnswer":83},"How do PURM and D3CL contribute during training, and what is kept for inference?",{"text":84,"@type":76},"PURM reconstructs the hazy observation from the restored result using transmission and atmospheric-light priors to enforce degradation consistency. D3CL regularizes clean restoration and hazy reconstruction spaces with real and synthetic negatives. During inference, only LPUM is retained, while PURM and D3CL are training-time constraints.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,127,130,134],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":22,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]