Hugging Face Daily Papers · · 6 min read

Decoupled Residual Denoising Diffusion Models for Unified and Data Efficient Image-to-Image Translation

Mirrored from Hugging Face Daily Papers for archival readability. Support the source by reading on the original site.

We propose Decoupled Residual Denoising Diffusion models (DRDD) for unified and data-efficient image-to-image (I2I) translation. While diffusion models have advanced I2I<br>translation in terms of quality and diversity, we uncover a previously under-explored property in diffusion models. Crucially, beyond its conventional role of manifold lifting<br>(i.e., moving data off low-dimensional manifolds), injecting<br>Gaussian noise facilitates domain harmonization by implicitly aligning feature distributions across domains, a property particularly advantageous for unified I2I translation.<br>However, existing diffusion models prematurely erode this<br>harmonization effect, as noise and residuals are simultaneously removed in a single coupled diffusion process. To address this, DRDD decouples the diffusion process into two<br>sequential and independent diffusion stages: (1) a stochastic noise diffusion for domain harmonization and manifold lifting, and (2) a deterministic residual diffusion that learns<br>the core semantic mapping entirely within the fixed-noise<br>domain. This decoupling preserves harmonization and manifold lifting effects throughout the transformation, substantially simplifying the learning of unified mappings across<br>diverse tasks and domains. Notably, the noise diffusion<br>stage is trained exclusively on abundant, unpaired targetdomain images, greatly improving data efficiency. Comprehensive theoretical and empirical analysis demonstrates that<br>DRDD is broadly compatible with mainstream diffusion models and consistently delivers robust, unified I2I translation,<br>even under limited paired data. Our code is available at<br><a href=\"https://github.com/HKU-HealthAI/DRDD\" rel=\"nofollow\">https://github.com/HKU-HealthAI/DRDD</a>.</p>\n","updatedAt":"2026-06-03T05:54:26.230Z","author":{"_id":"663058bc2653ec94f4a6235f","avatarUrl":"/avatars/f55b8c3c8100d6b6d65ba61abc4fb014.svg","fullname":"Liangqiong Qu","name":"Liangqiong-QU","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.859747052192688},"editors":["Liangqiong-QU"],"editorAvatarUrls":["/avatars/f55b8c3c8100d6b6d65ba61abc4fb014.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.01048","authors":[{"_id":"6a1fc138e292c1c78ecb1570","name":"Ziyue Lin","hidden":false},{"_id":"6a1fc138e292c1c78ecb1571","name":"Jiahe Hou","hidden":false},{"_id":"6a1fc138e292c1c78ecb1572","name":"Hongyu Xia","hidden":false},{"_id":"6a1fc138e292c1c78ecb1573","name":"Xinrui Xie","hidden":false},{"_id":"6a1fc138e292c1c78ecb1574","name":"Feifei Wang","hidden":false},{"_id":"6a1fc138e292c1c78ecb1575","name":"Yuyin Zhou","hidden":false},{"_id":"6a1fc138e292c1c78ecb1576","name":"Wei Wang","hidden":false},{"_id":"6a1fc138e292c1c78ecb1577","name":"Jiawei Liu","hidden":false},{"_id":"6a1fc138e292c1c78ecb1578","name":"Liangqiong Qu","hidden":false}],"publishedAt":"2026-05-31T06:38:18.000Z","submittedOnDailyAt":"2026-06-03T00:00:00.000Z","title":"Decoupled Residual Denoising Diffusion Models for Unified and Data Efficient Image-to-Image Translation","submittedOnDailyBy":{"_id":"663058bc2653ec94f4a6235f","avatarUrl":"/avatars/f55b8c3c8100d6b6d65ba61abc4fb014.svg","isPro":false,"fullname":"Liangqiong Qu","user":"Liangqiong-QU","type":"user","name":"Liangqiong-QU"},"summary":"We propose Decoupled Residual Denoising Diffusion models (DRDD) for unified and data-efficient image-to-image (I2I) translation. While diffusion models have advanced I2I translation in terms of quality and diversity, we uncover a previously under-explored property in diffusion models. Crucially, beyond its conventional role of manifold lifting (i.e., moving data off low-dimensional manifolds), injecting Gaussian noise facilitates domain harmonization by implicitly aligning feature distributions across domains, a property particularly advantageous for unified I2I translation. However, existing diffusion models prematurely erode this harmonization effect, as noise and residuals are simultaneously removed in a single coupled diffusion process. To address this, DRDD decouples the diffusion process into two sequential and independent diffusion stages: (1) a stochastic noise diffusion for domain harmonization and manifold lifting, and (2) a deterministic residual diffusion that learns the core semantic mapping entirely within the fixed-noise domain. This decoupling preserves harmonization and manifold lifting effects throughout the transformation, substantially simplifying the learning of unified mappings across diverse tasks and domains. Notably, the noise diffusion stage is trained exclusively on abundant, unpaired target-domain images, greatly improving data efficiency. Comprehensive theoretical and empirical analysis demonstrates that DRDD is broadly compatible with mainstream diffusion models and consistently delivers robust, unified I2I translation, even under limited paired data. Our code is available at https://github.com/HKU-HealthAI/DRDD.","upvotes":10,"discussionId":"6a1fc138e292c1c78ecb1579","githubRepo":"https://github.com/HKU-HealthAI/DRDD","githubRepoAddedBy":"user","ai_summary":"Decoupled Residual Denoising Diffusion models (DRDD) improve unified image-to-image translation by separating noise diffusion for domain harmonization from residual diffusion for semantic mapping, enhancing data efficiency and performance.","ai_keywords":["diffusion models","image-to-image translation","domain harmonization","noise diffusion","residual diffusion","manifold lifting","unified I2I translation","data efficiency"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":6,"organization":{"_id":"67ea9ecfc234715db8dbf339","name":"hkuhk","fullname":"The University of Hong Kong","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/67ea9e8d2d95c10a0da11b0c/FNnR4M7YqKRuG43N5771B.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"663058bc2653ec94f4a6235f","avatarUrl":"/avatars/f55b8c3c8100d6b6d65ba61abc4fb014.svg","isPro":false,"fullname":"Liangqiong Qu","user":"Liangqiong-QU","type":"user"},{"_id":"66632d290875aaaa914e7335","avatarUrl":"/avatars/07185efaa8a170f555ebf3eeef88bb6c.svg","isPro":false,"fullname":"Zonggen Li","user":"ZonggenLi","type":"user"},{"_id":"6852b9610e0d43341926b9a2","avatarUrl":"/avatars/6d672f8830e1453cdc5a4686ce252bbd.svg","isPro":false,"fullname":"junfutan","user":"jun0519","type":"user"},{"_id":"668f440894dfc0ed1a7006ed","avatarUrl":"/avatars/fa0d328300b03bcbbf9b3a7532f28458.svg","isPro":false,"fullname":"Pengxin Guo","user":"gpx333","type":"user"},{"_id":"6440f16a5d600fb095198f15","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6440f16a5d600fb095198f15/-J89Lelh3PPRVcc1ya_VS.png","isPro":false,"fullname":"Jiawei Liu","user":"nachifur","type":"user"},{"_id":"66ff619fe48de0216cd43531","avatarUrl":"/avatars/e4642e02b6475cfbd677c6e28640b5b0.svg","isPro":false,"fullname":"HaoningJiang","user":"haoning666","type":"user"},{"_id":"67dcd54539fdee91f9661e6b","avatarUrl":"/avatars/6bd2b52b5ed66f29a100cdb683d7fa81.svg","isPro":false,"fullname":"Hou JIahe","user":"S462255048","type":"user"},{"_id":"68fa25bddc4736cd8614b80d","avatarUrl":"/avatars/7c6b8b79749e1dcc40ca2cb2d9a975c4.svg","isPro":false,"fullname":"Xia","user":"XHYu","type":"user"},{"_id":"6807d368391ae636b489cf84","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/8FdoZ3f08DGpbAvZGywx2.png","isPro":false,"fullname":"XHYu","user":"XYuuuu","type":"user"},{"_id":"6810c14f740c4365f349ca6e","avatarUrl":"/avatars/9a4de12a354764f0ff740e8e77f0181d.svg","isPro":false,"fullname":"Runxi Wang","user":"1wrx1","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"67ea9ecfc234715db8dbf339","name":"hkuhk","fullname":"The University of Hong Kong","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/67ea9e8d2d95c10a0da11b0c/FNnR4M7YqKRuG43N5771B.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.01048.md"}">
Papers
arxiv:2606.01048

Decoupled Residual Denoising Diffusion Models for Unified and Data Efficient Image-to-Image Translation

Published on May 31
· Submitted by
Liangqiong Qu
on Jun 3
Authors:
,
,
,
,
,
,
,
,

Abstract

Decoupled Residual Denoising Diffusion models (DRDD) improve unified image-to-image translation by separating noise diffusion for domain harmonization from residual diffusion for semantic mapping, enhancing data efficiency and performance.

We propose Decoupled Residual Denoising Diffusion models (DRDD) for unified and data-efficient image-to-image (I2I) translation. While diffusion models have advanced I2I translation in terms of quality and diversity, we uncover a previously under-explored property in diffusion models. Crucially, beyond its conventional role of manifold lifting (i.e., moving data off low-dimensional manifolds), injecting Gaussian noise facilitates domain harmonization by implicitly aligning feature distributions across domains, a property particularly advantageous for unified I2I translation. However, existing diffusion models prematurely erode this harmonization effect, as noise and residuals are simultaneously removed in a single coupled diffusion process. To address this, DRDD decouples the diffusion process into two sequential and independent diffusion stages: (1) a stochastic noise diffusion for domain harmonization and manifold lifting, and (2) a deterministic residual diffusion that learns the core semantic mapping entirely within the fixed-noise domain. This decoupling preserves harmonization and manifold lifting effects throughout the transformation, substantially simplifying the learning of unified mappings across diverse tasks and domains. Notably, the noise diffusion stage is trained exclusively on abundant, unpaired target-domain images, greatly improving data efficiency. Comprehensive theoretical and empirical analysis demonstrates that DRDD is broadly compatible with mainstream diffusion models and consistently delivers robust, unified I2I translation, even under limited paired data. Our code is available at https://github.com/HKU-HealthAI/DRDD.

Community

We propose Decoupled Residual Denoising Diffusion models (DRDD) for unified and data-efficient image-to-image (I2I) translation. While diffusion models have advanced I2I
translation in terms of quality and diversity, we uncover a previously under-explored property in diffusion models. Crucially, beyond its conventional role of manifold lifting
(i.e., moving data off low-dimensional manifolds), injecting
Gaussian noise facilitates domain harmonization by implicitly aligning feature distributions across domains, a property particularly advantageous for unified I2I translation.
However, existing diffusion models prematurely erode this
harmonization effect, as noise and residuals are simultaneously removed in a single coupled diffusion process. To address this, DRDD decouples the diffusion process into two
sequential and independent diffusion stages: (1) a stochastic noise diffusion for domain harmonization and manifold lifting, and (2) a deterministic residual diffusion that learns
the core semantic mapping entirely within the fixed-noise
domain. This decoupling preserves harmonization and manifold lifting effects throughout the transformation, substantially simplifying the learning of unified mappings across
diverse tasks and domains. Notably, the noise diffusion
stage is trained exclusively on abundant, unpaired targetdomain images, greatly improving data efficiency. Comprehensive theoretical and empirical analysis demonstrates that
DRDD is broadly compatible with mainstream diffusion models and consistently delivers robust, unified I2I translation,
even under limited paired data. Our code is available at
https://github.com/HKU-HealthAI/DRDD.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images

· Sign up or log in to comment

Get this paper in your agent:

hf papers read 2606.01048
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2606.01048 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2606.01048 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2606.01048 in a Space README.md to link it from this page.

Collections including this paper 0

No Collection including this paper

Add this paper to a collection to link it from this page.

Discussion (0)

Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.

Sign in →

No comments yet. Sign in and be the first to say something.

More from Hugging Face Daily Papers