[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-83935-en":3,"doc-seo-83935-105":29,"detail-sidebar-cat-0-en-105":90},{"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":4,"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},83935,1099514068035,"Ezra","https://ap-avatar.wpscdn.com/davatar_276721f389ce27ea32af1340a28f341c",8,"Research & Report","FlowMark Mask Guided Video Watermarking","FlowMark presents a mask-guided framework for video watermarking that uses automatically predicted object masks to select embedding regions without manual guidance. An end-to-end, region-aware architecture is trained with noise augmentation to preserve perceptual quality while improving robustness to compression, geometric distortions, and content variation. Content-adaptive masking maintains coherent watermark signals and eliminates perceptual flicker across frames. FlowMark also supports reliable recovery under video-native temporal edits and social-media re-encoding, embedding 128-bit messages with strong PSNR for provenance and integrity protection.","arXiv :2607 .0526 1v 1 [ cs .CV] 6 Jul 2026  \nFlowMark: Mask-Guided Video Watermarking  \nVishal Asnani 1 , Shruti Agarwal 1 , and John Collomosse 1 ,2  \n1 Adobe Research, 2 DECaDE, University of Surrey  \n{vasnani, shragarw, [collomos}@adobe.com](collomos}@adobe.com)  \nAbstract. We present FlowMark, a video watermarking framework guided by automatically predicted object masks. In contrast to prior region-based approaches that require user-supplied mask guidance, FlowMark learns to identify optimal regions for watermark embedding through a dedicated Mask Predictor network. Our end-to-end trainable architecture combines region-aware encoding with noise-augmented training to ensure robustness against compression, geometric transformations, and content variation, while preserving high perceptual quality.  \nOur content-adaptive masking keeps watermark signals coherent with natural video dynamics, effectively eliminating perceptual flicker. Beyond compression robustness, FlowMark maintains reliable watermark recovery under video-native temporal edits (e.g. , frame swap, insertion, deletion, resampling, and interpolation) and real-world social media distribution pipelines (e.g., YouTube and Facebook re-encoding) . Experimental results on both image and video datasets show that FlowMark reliably embeds 128-bit messages with up to 50.08 dB PSNR, offering strong performance for content provenance, temporal authenticity verification, and video integrity protection.  \n1 Introduction  \nAdvances in generative video are enabling highly realistic synthetic content, unlocking powerful creative tools but also raising concerns around misuse. A common mitigation is to attach provenance metadata that raises awareness of content origins (e.g., Sora and Veo3 attach C2PA metadata [17]) . For images, such metadata is often reinforced through invisible watermarking [18, 27], embedding durable identifiers that persist through format conversion and content platforms.  \nExtending watermarking to video presents unique challenges beyond those encountered in image watermarking. Most existing approaches rely on global spatial or frequency-domain embedding [14,62], and when applied independently to video frames, they often introduce temporal inconsistencies such as flicker due toa lack of coherence across frames. These artifacts become especially pronounced under the spatio-temporal distortions introduced by modern video codecs like H.264 and H.265, resulting in fragile watermark recovery. While recent techniques such as VideoSeal [16] are trained for robustness against compression, they still suffer from visible flickering because the embedded watermark can vary from frame to frame. Other approaches like WAM [43] and MaskWM [31]  \n2 Asnani et al.  \nFig. 1: (a) VideoSeal [23] embeds watermarks across the full frame, WAM [43] uses segmentation masks, and MaskWM [30] relies on random masks. FlowMark learns content-aware masks that place the watermark in perceptually resilient regions. The residual R is the pixel-wise difference between the original and watermarked frames, and a lower ∥R∥1 indicates less visible change and reduced flickering across consecutive video frames. FlowMark achieves the lowest residual norm among all methods. (b) FlowMark achieves highest watermark detection robustness on social media websites.  \nincorporate Just Noticeable Difference (JND) models [52] and per-frame region masking to improve perceptual quality, but these require manual mask selection during encoding, a process that is impractical for real-world video pipelines. These approaches embed signals globally or within fixed segmentation or random masks, often leading to uneven watermark energy and higher residual L 1 norms across frames ( Fig. 1) . A related strategy for mitigating flicker in neural watermarking is to embed the same watermark message across multiple consecutive frames [23] . Although this stabilizes appearance, it inherently restricts frame-level uniqueness, prev","cbCaigHYjbrUAfpL","https://ap.wps.com/l/cbCaigHYjbrUAfpL","pdf",17882135,1,36,"English","en",105,"# Introduction\n## Challenges in video watermarking\n## Proposed approach: FlowMark\n## Training strategy and robustness evaluation","[{\"question\":\"What problem does FlowMark address in existing video watermarking methods?\",\"answer\":\"FlowMark targets temporal inconsistencies such as perceptual flicker caused when watermark signals are applied independently across video frames, especially under codec-induced spatio-temporal distortions.\"},{\"question\":\"How does FlowMark generate watermark embedding regions without manual masks?\",\"answer\":\"FlowMark uses a dedicated Mask Predictor network to automatically learn optimal spatial regions for watermark embedding, making the framework end-to-end trainable.\"},{\"question\":\"What kinds of attacks or transformations does FlowMark remain robust against?\",\"answer\":\"FlowMark is trained and evaluated for robustness against compression, geometric transformations, and video-native temporal edits such as frame reordering, insertion, deletion, resampling, and interpolation, including social-media re-encoding pipelines.\"}]",1784191541,91,{"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":85,"head_meta":87,"extra_data":89,"updated_unix":27},"flowmark-mask-guided-video-watermarking","",{"@graph":35,"@context":84},[36,53,67],{"@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/flowmark-mask-guided-video-watermarking/83935/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"What problem does FlowMark address in existing video watermarking methods?","Question",{"text":74,"@type":75},"FlowMark targets temporal inconsistencies such as perceptual flicker caused when watermark signals are applied independently across video frames, especially under codec-induced spatio-temporal distortions.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How does FlowMark generate watermark embedding regions without manual masks?",{"text":79,"@type":75},"FlowMark uses a dedicated Mask Predictor network to automatically learn optimal spatial regions for watermark embedding, making the framework end-to-end trainable.",{"name":81,"@type":72,"acceptedAnswer":82},"What kinds of attacks or transformations does FlowMark remain robust against?",{"text":83,"@type":75},"FlowMark is trained and evaluated for robustness against compression, geometric transformations, and video-native temporal edits such as frame reordering, insertion, deletion, resampling, and interpolation, including social-media re-encoding pipelines.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":91},[92,96,100,104,109,114,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":105,"slug":137},19,"General","general"]