[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-82288-en":3,"doc-seo-82288-105":29,"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":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},82288,13056703019662,"Evangeline","https://ap-avatar.wpscdn.com/avatar/be000253a8e92610077?_k=1778726343310543188",8,"Research & Report","Rethinking Monocular Depth Embedding for Generalized Stereo Matching","The document addresses generalized stereo matching by combining monocular depth priors with stereo geometric constraints, aiming to improve dense disparity estimation across diverse scenes. It identifies two core issues: unstable alignment between monocular and stereo cues and performance degradation from unreliable monocular signals. The proposed approach reduces branch coupling to avoid shortcut learning, replaces hard constraints with soft constraints from monocular depth, fuses monocular depth with RGB, guides iterative GRU disparity updates using depth gradients, and adds edge-confidence estimation with an edge-aware loss to handle boundary blur from augmentation. Experiments report state-of-the-art generalization.","arXiv :2607 .09284v1 [ cs .CV] 10 Jul 2026  \nRethinking Monocular Depth Embedding for Generalized Stereo Matching  \nLibo Lina , Shuangli Dua,∗, Minghua Zhaoa , Zhenzhen Youa , Shun Lvb , Yiguang Liub  \na Shaanxi Key Laboratory for Network Computing and Security Technology, Xi’an University of Technology, No. 5, Jinhua South  \nRoad, Xi’an, 710048, Shaanxi, China  \nb College of Computer Science, Sichuan University,, Chengdu, 610065, Sichuan, China  \nAbstract  \nGenerally, monocular methods capture rich contextual priors but lack geometric precision, whereas stereo methods are geometrically accurate yet struggle in textureless and occluded regions. Several approaches attempt to combine their strengths to enhance the generalization of stereo matching (SM) by aligning monocular depth with stereo information. However, establishing a stable and generalizable alignment is challenging, and unreliable monocular cues can substantially degrade performance. This paper rethinks monocular depth embedding. First, to prevent shortcut learning, we reduce branch coupling instead of expanding network width. Second, we construct soft constraints instead of hard ones from monocular depth to improve tolerance to monocular depth errors. Based on the principles, we integrate monocular information into both feature extraction and GRU iterations. Specifically, the monocular depth map is fused with the RGB image to sharpen depth boundary perception and suppress matching ambiguities. The fused image is then used for feature extraction, allowing the contextual features to encode global geometric information. Furthermore, the monocular depth gradient feature is employed to guide disparity updates, helping to escape local oscillations. Finally, to address the boundary blurring of supervised disparity caused by data augmentation, we propose an edge confidence estimation method and an edge-aware loss function. Our method achieves state-of-theart (SOTA) performance on multiple standard benchmarks, demonstrating excellent generalization while improving accuracy. The code is available at [https:](https://github.com/linliboabc-maker/stereo-matching-digital)[//](https://github.com/linliboabc-maker/stereo-matching-digital)[github.com](https://github.com/linliboabc-maker/stereo-matching-digital)[/](https://github.com/linliboabc-maker/stereo-matching-digital)[linliboabc-maker](https://github.com/linliboabc-maker/stereo-matching-digital)[/](https://github.com/linliboabc-maker/stereo-matching-digital)[stereo-matching-digital](https://github.com/linliboabc-maker/stereo-matching-digital).  \nKeywords: stereo matching, monocular depth, iterative optimization, edge detection  \n1. Introduction  \nStereo matching (SM) aims to estimate dense, pixelwise disparity maps from a pair of rectified stereo images [1, 2] . The resulting disparity maps can subsequently be converted into metric depth for various downstream applications, including autonomous driving, robotics, augmented reality, etc. These applications require that SM methods generalize effectively across diverse real-world scenes [3] .  \nA typical SM pipeline [4, 1] includes four stages: feature extraction, cost volume generation, cost aggregation, and disparity regression. After embedding these steps into an end-to-end deep network [5], many deep learning-based SM methods emerged with significantly improved performance. Early deep SM primarily relied on filtering-based techniques to refine the cost volume and subsequently regress accurate disparities. These approaches typically begin by constructing a 3D or 4D  \n∗ Corresponding author. Email: [dusl@xaut.edu.cn](dusl@xaut.edu.cn)  \ncost volume, which is then filtered using 2D or 3D convolutional neural networks (CNNs)[4, 6, 7, 8] . However, the single-stage regression makes the learned mapping highly dependent on training data. Then, any domain shift (e.g., changes in lighting, texture, or noise) attest time will cause severe performance degradation.  \nLater, following the ","cbCaiuftgeQVjXa6","https://ap.wps.com/l/cbCaiuftgeQVjXa6","pdf",20793827,1,15,"English","en",105,"# Introduction\n# Method Overview\n## Monocular Depth Embedding Design\n## Feature Extraction and Fusion\n## Iterative Disparity Update with GRU\n## Edge Confidence and Edge-Aware Loss\n# Experiments and Results","[{\"question\":\"Why is combining monocular depth with stereo matching challenging for generalized stereo matching?\",\"answer\":\"Monocular priors provide context but can be geometrically imprecise, while stereo is geometric yet fails in ambiguous regions. Achieving stable, generalizable alignment is difficult, and unreliable monocular cues can noticeably degrade performance.\"},{\"question\":\"How does the method prevent shortcut learning when embedding monocular depth?\",\"answer\":\"Instead of widening the network, it reduces branch coupling to avoid learning shortcut behaviors that harm generalization.\"},{\"question\":\"What mechanisms improve boundary accuracy and reduce blur from data augmentation?\",\"answer\":\"An edge confidence estimation method is introduced alongside an edge-aware loss function, improving supervised disparity boundaries under augmentation-induced boundary blurring.\"}]",1784179412,38,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"rethinking-monocular-depth-embedding-for-generalized-stereo-matching","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/rethinking-monocular-depth-embedding-for-generalized-stereo-matching/82288/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","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 combining monocular depth with stereo matching challenging for generalized stereo matching?","Question",{"text":75,"@type":76},"Monocular priors provide context but can be geometrically imprecise, while stereo is geometric yet fails in ambiguous regions. Achieving stable, generalizable alignment is difficult, and unreliable monocular cues can noticeably degrade performance.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the method prevent shortcut learning when embedding monocular depth?",{"text":80,"@type":76},"Instead of widening the network, it reduces branch coupling to avoid learning shortcut behaviors that harm generalization.",{"name":82,"@type":73,"acceptedAnswer":83},"What mechanisms improve boundary accuracy and reduce blur from data augmentation?",{"text":84,"@type":76},"An edge confidence estimation method is introduced alongside an edge-aware loss function, improving supervised disparity boundaries under augmentation-induced boundary blurring.","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":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]