[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86171-en":3,"doc-seo-86171-105":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},86171,962075114101,"Seraphina","https://ap-avatar.wpscdn.com/avatar/e000253a75eb197efd?x-image-process=image/resize,m_fixed,w_180,h_180&k=1780044092746381165",8,"Research & Report","Parallax Portrait Matting","Image matting is an ill-posed decomposition that becomes especially difficult when foreground and background contain rich texture. Single-image approaches often rely on learned priors and still fail on challenging portraits. Parallax Portrait Matting introduces a practical two-frame method that uses a second image with slight viewpoint change, typical of burst photography, to provide complementary constraints. The pipeline estimates trimaps and foreground/background motion, aligns views, and fuses information conservatively to recover finer details and more accurate foreground colors.","arXiv :2607 . 11205v1 [ cs .CV] 13 Jul 2026  \nParallax Portrait Matting⋆  \nXin Cai 1 ,3 , Jiawen Chen2 , Lars Jebe2 , Tianfan Xue 1 ,3 ,4 , and Zhoutong Zhang2†  \n1 Multimedia Laboratory, The Chinese University of Hong Kong, Hong Kong SAR, China  \n2 Adobe NextCam, San Jose, CA, USA  \n3 Shanghai AI Laboratory, Shanghai, China  \n4 CPII under InnoHK, Hong Kong SAR, China  \nAbstract. Image matting is highly ill-posed, especially when both the foreground and background are richly textured. While single-image matting methods learn strong priors from data, they often struggle on these challenging cases. Existing approaches improve results by requiring additional signals such as green screens, polarized lighting, or clean background images, but these typically rely on specialized capture setups.  \nWe present Parallax Portrait Matting, a practical two-frame matting method that uses a second image captured with slight viewpoint change.  \nSuch a setting arises naturally in burst photography, where small camera motion induces foreground-background parallax and provides complementary observations for matting. Our pipeline estimates trimaps and foreground/background motion, then constructs aligned views for prediction. To handle imperfect motion estimation, the network uses the background-aligned pair for direct fusion and the foreground-aligned cue through cross-attention for error compensation. Experiments show that our method recovers finer details and more accurate foreground colors than strong single-image matting baselines on challenging portrait cases.  \n1 Introduction  \nImage matting has a long history in computer vision and graphics [10, 31] . This decades-old problem aims to decompose an image I into a foreground image F , a background image B , and an opacity (alpha) map α, where they jointly reconstruct the input image through a linear composition process:  \nI = αF + (1 − α)B.  \nLike most inverse problems, the matting problem is known to be ill-posed. To arrive at a solution representing the actual scene, it requires either priors over F , B , and α, or additional information as constraints to the solution space.  \nMost modern methods resolve this ambiguity either by learning strong priors from annotated data [21, 27 , 41 , 44 , 45] or by leveraging generative models [38], but single-image prediction remains difficult in highly ambiguous cases. Another  \n⋆ Project page: [https://caixin98.github.io/parallax/](https://caixin98.github.io/parallax/)  \n2 X. Cai et al.  \nTwo Input frame with parallax  \n(a) Base Frame from Input (b) Foreground & alpha, remove.bg (c) Foreground & alpha, ours  \nFig. 1: Our matting method exploits camera-motion-induced parallax between the foreground and the background. It takes two frames as input, each taken with a slightly different camera location, and predicts both a pre-multiplied foreground image and an alpha map. Trained on public datasets, our method produces a cleaner foreground with more details than closed-source commercial solutions like [remove.bg](remove.bg).  \nline of work tackles the ill-posedness of single-image matting by acquiring additional observations. However, existing solutions often rely on specialized capture setups such as green screens, polarization, camera arrays, focal stacks, or clean background images [1, 2 , 9 , 13 , 14 , 29] . These approaches can produce high-quality mattes, especially for challenging regions such as hair and semi-transparent boundaries, but they typically require dedicated hardware or carefully controlled capture procedures, making them impractical for everyday photography.  \nIn this work, we introduce Parallax Portrait Matting (Fig. 1), a practical two-frame matting method for portrait scenes. Our key idea is to exploit an additional image captured with slight viewpoint change, which is often already available in casual photography or mobile burst capture. Because the portrait subject is typically closer to the camera than the background, the two laye","cbCaijfG57g2WZYf","https://ap.wps.com/l/cbCaijfG57g2WZYf","pdf",10267957,2,1,18,"English","en",105,"# Introduction\n## Image matting as an ill-posed inverse problem\n## Single-image limitations and prior-based or generative approaches\n## Parallax Portrait Matting: core idea and practical motivation\n## Motion modeling challenges and conservative fusion strategy","[{\"question\":\"What is the main problem addressed by Parallax Portrait Matting?\",\"answer\":\"Image matting is ill-posed, particularly for portraits with richly textured foreground and background, where single-image methods struggle to produce accurate alpha and foreground colors.\"},{\"question\":\"How does the method obtain extra information beyond a single image?\",\"answer\":\"It uses a second frame captured with slight viewpoint change, which naturally occurs in burst photography and introduces foreground–background parallax.\"},{\"question\":\"How does the network handle unreliable motion estimation in difficult regions?\",\"answer\":\"It models motion conservatively: background-aligned inputs are used for direct fusion, while foreground-aligned cues are introduced via cross-attention in feature space to compensate for residual alignment errors.\"}]",1784209095,45,{"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},"parallax-portrait-matting","",{"@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/parallax-portrait-matting/86171/",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},"What is the main problem addressed by Parallax Portrait Matting?","Question",{"text":75,"@type":76},"Image matting is ill-posed, particularly for portraits with richly textured foreground and background, where single-image methods struggle to produce accurate alpha and foreground colors.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the method obtain extra information beyond a single image?",{"text":80,"@type":76},"It uses a second frame captured with slight viewpoint change, which naturally occurs in burst photography and introduces foreground–background parallax.",{"name":82,"@type":73,"acceptedAnswer":83},"How does the network handle unreliable motion estimation in difficult regions?",{"text":84,"@type":76},"It models motion conservatively: background-aligned inputs are used for direct fusion, while foreground-aligned cues are introduced via cross-attention in feature space to compensate for residual alignment errors.","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},[93,97,101,105,110,115,120,123,128,131,135],{"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":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]