[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-156469-105":59,"doc-detail-156469-en":130},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":123,"head_meta":125,"extra_data":127,"updated_unix":129},105,"en","fsi-edit-frequency-and-stochasticity-injection-for-flexible-diffusion-based-image-editing","FSI-Edit - Frequency and Stochasticity Injection for Flexible Diffusion-Based Image Editing","","Latent Diffusion-based Text-to-Image (T2I) editing typically inverts an image into noise, reconstructs it using the original prompt, then generates an edited result under a new target prompt. Directly injecting reconstruction features to preserve unchanged content can create feature incompatibility, reducing editing fidelity and restricting creative flexibility, especially for non-rigid edits like pose or structural changes. FSI-Edit introduces frequency-aware high-frequency residual injection for better consistency and controlled stochasticity in replaced features to enlarge the generative space. Experiments on non-rigid additions, deletions, and pose manipulation show improved alignment, semantic fidelity, and visual quality, highlighting frequency and stochasticity as key to reducing rigidity.",{"@graph":69,"@context":122},[70,84,105],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/fsi-edit-frequency-and-stochasticity-injection-for-flexible-diffusion-based-image-editing/156469/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/fsi-edit-frequency-and-stochasticity-injection-for-flexible-diffusion-based-image-editing/156469.png","ImageObject",300,407,{"name":92,"@type":93},"Olivia Brown","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-09-17","2026-08-28",true,{"@type":102,"interactionType":103,"userInteractionCount":81},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"What problem does FSI-Edit address in diffusion-based image editing?","Question",{"text":112,"@type":113},"Direct feature replacement from reconstruction to generation often causes feature incompatibility, which harms editing fidelity and limits creative flexibility, particularly for non-rigid edits such as pose or structural changes.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"How does FSI-Edit improve feature consistency during editing?",{"text":117,"@type":113},"FSI-Edit injects high-frequency components from reconstruction features into generation features, reducing incompatibility while preserving the editing ability for major structures encoded in low-frequency information.",{"name":119,"@type":110,"acceptedAnswer":120},"How does FSI-Edit enable more diverse non-rigid edits?",{"text":121,"@type":113},"It adds controlled noise into the replaced reconstruction features, expanding the generative space so the model can produce diverse non-rigid transformations beyond constraints of the original image.","https://schema.org",{"og:url":83,"og:type":124,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":126,"canonical":83},"index,follow",{"doc_id":128,"site_id":62},156469,1787959498,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":81,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":139,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":129,"read_time":144},16904993612988,"https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd","FSI-Edit: Frequency and Stochasticity Injection for Flexible Diffusion-Based Image Editing  \nKaixiang Yang†, Xin Li†, Yuxi Li†, Qiang Li, Zhiwei Wang∗  \nWuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology  \n† : Co-first authors, ∗ : Corresponding author.  \n{kxyang, lixin2023, liyuxi9, liqiang8, [zwwang}@hust.edu.cn](zwwang}@hust.edu.cn)  \nNon-Rigid Editing Rigid Editing  \n\n| | |\n| --- | --- |\n\nFigure 1: Our method is capable of handling non-rigid editing, including pose changes, object addition, and removal, as well as rigid editing.  \nAbstract  \nLatent Diffusion-based Text-to-Image (T2I) is a free image editing tool that typically reverses an image into noise, reconstructs it using its original text prompt, and then generates an edited version under a new target prompt. To preserve unaltered image content, features from the reconstruction are directly injected to replace selected features in the generation. However, this direct replacement often leads to feature incompatibility, compromising editing fidelity and limiting creative flexibility, particularly for non-rigid edits (e.g., structural or pose changes) . In this paper, we aim to address these limitations by proposing FSI-Edit, a novel framework using frequency-and stochasticity-based feature injection for flexible image editing.  \nFirst, FSI-Edit enhances feature consistency by injecting high-frequency components of reconstruction features into generation features, mitigating incompatibility while preserving the editing ability for major structures encoded in low-frequency information. Second, it introduces controlled noise into the replaced reconstruction features, expanding the generative space to enable diverse non-rigid edits beyond the original image’s constraints. Experiments on non-rigid edits, e.g., addition, deletion, and pose manipulation, demonstrate that FSI-Edit outperforms existing baselines in target alignment, semantic fidelity and visual quality. Our work highlights the critical roles of frequency-aware design and stochasticity in overcoming rigidity in diffusion-based editing.  \n39th Conference on Neural Information Processing Systems (NeurIPS 2025) .  \nTarget Generation Source Reconstruction  \nA sitting dog  \nInversed xT × T  \n❄  \n Ajumping dog   \n(a) Pure T2I  \nA sitting dog  \n  ❄  \nInversed xT  \n× T  \n❄  \ngap  \nInversed xT  \n(b) Existing Image Editing  \nFigure 2: Editing Paradigm Comparison. Top: source reconstruction; Bottom: target generation. (a) Pure T2I: there is no interaction between source and target images, results are random and unrelated to the source.(b) Existing Image Editing: typical editing methods directly inject features, causing semantic gaps and low flexibility, especially for non-rigid edits. (c) Ours: injects only high-frequency residuals and adds stochasticity, reducing semantic gap and improving edit quality.  \n1 Introduction  \nDiffusion models [1] have achieved remarkable success in the domain of Text-to-Image (T2I) generation in recent advances [2–7] . In particular, Latent Diffusion Models (LDMs) demonstrate exceptional ability to translate textual descriptions (i.e., prompts) into high-quality images, leading to their widespread adoption in downstream applications such as image [8, 9] and video editing [10, 11] . Image editing refers to the task of transforming a source image into a desired target image guided by user-provided prompts. This technology has become an integral part of daily life, with broad applications across domains such as social media and visual effects.  \nThe typical workflow of LDM-based image editing follows a common paradigm. First, the source image is inverted back into a latent noise representation, typically through DDIM inversion [12] or more advanced strategies [13–15] . From this latent noise, two parallel denoising processes are initiated: one reconstructs the original image conditioned on the source prompt, while the other generates a modified version ","cbCaikDAhE4O3xIO","https://ap.wps.com/l/cbCaikDAhE4O3xIO","pdf",20357132,33,"English","# Abstract\n# Introduction\n## Background: Latent diffusion and inversion\n## Existing feature replacement approaches\n## Challenges in non-rigid editing\n## Proposed direction: frequency-aware and stochasticity-based injection","[{\"question\":\"What problem does FSI-Edit address in diffusion-based image editing?\",\"answer\":\"Direct feature replacement from reconstruction to generation often causes feature incompatibility, which harms editing fidelity and limits creative flexibility, particularly for non-rigid edits such as pose or structural changes.\"},{\"question\":\"How does FSI-Edit improve feature consistency during editing?\",\"answer\":\"FSI-Edit injects high-frequency components from reconstruction features into generation features, reducing incompatibility while preserving the editing ability for major structures encoded in low-frequency information.\"},{\"question\":\"How does FSI-Edit enable more diverse non-rigid edits?\",\"answer\":\"It adds controlled noise into the replaced reconstruction features, expanding the generative space so the model can produce diverse non-rigid transformations beyond constraints of the original image.\"}]","FSI-Edit - Frequency and Stochasticity Injection for Flexible Diffusion-Based Image Editing | PDF",83]