One Noise to Rule Them All: Learning a Unified Model of Spatially-Varying Noise Patterns
Not on the map. Why?
Abstract (arXiv)
Procedural noise is a fundamental component of computer graphics pipelines, offering a flexible way to generate textures that exhibit "natural" random variation. Many different types of noise exist, each produced by a separate algorithm. In this paper, we present a single generative model which can learn to generate multiple types of noise as well as blend between them. In addition, it is capable of producing spatially-varying noise blends despite not having access to such data for training. These features are enabled by training a denoising diffusion model using a novel combination of data augmentation and network conditioning techniques. Like procedural noise generators, the model's behavior is controllable via interpretable parameters and a source of randomness. We use our model to produce a variety of visually compelling noise textures. We also present an application of our model to improving inverse procedural material design; using our model in place of fixed-type noise nodes in a procedural material graph results in higher-fidelity material reconstructions without needing to know the type of noise in advance.
Key facts. Cited by 2 papers in the map (data through 13 Oct 2026). Builds on 22 of its 59 references that are in the map. First posted on arXiv 25 Apr 2024.
Builds on
22 of 59 references.
References · 22
- 2020Denoising Diffusion Probabilistic Modelscited by 20,952
- 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,082
- 2020Denoising Diffusion Implicit Modelscited by 9,718
- 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,213
- 2021Improved Denoising Diffusion Probabilistic Modelscited by 3,751
- 2019CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Featurescited by 2,882
- 2018Group Normalizationcited by 1,518
- 2019Semantic Image Synthesis with Spatially-Adaptive Normalizationcited by 1,296
- 2023Common Diffusion Noise Schedules and Sample Steps are Flawedcited by 309
- 2016Precomputed Real-Time Texture Synthesis with Markovian Generative Adversarial Networkscited by 192
- 2016Texture Networks: Feed-forward Synthesis of Textures and Stylized Imagescited by 173
- 2020A U-Net Based Discriminator for Generative Adversarial Networkscited by 128
10 more
- 2020UCSG-Net -- Unsupervised Discovering of Constructive Solid Geometry Treecited by 77
- 2017Learning Texture Manifolds with the Periodic Spatial GANcited by 53
- 2021CSG-Stump: A Learning Friendly CSG-Like Representation for Interpretable Shape Parsingcited by 51
- 2021DU-GAN: Generative Adversarial Networks with Dual-Domain U-Net Based Discriminators for Low-Dose CT Denoisingcited by 46
- 2023Don't Play Favorites: Minority Guidance for Diffusion Modelscited by 37
- 2022SeamlessGAN: Self-Supervised Synthesis of Tileable Texture Mapscited by 19
- 2023PhotoMat: A Material Generator Learned from Single Flash Photoscited by 18
- 2017A survey of exemplar-based texture synthesiscited by 13
- 2022Node Graph Optimization Using Differentiable Proxiescited by 13
- 2019A Bayesian Inference Framework for Procedural Material Parameter Estimationcited by 11
Built on by
Citing papers per year: 2024: 1; 2025: 1.
2 papers.
Similar, not cited
Close in wording, no citation either way
- 2024GenesisTex: Adapting Image Denoising Diffusion to Texture Space
- 2026Toward Richer Material Generation via Procedural Data Enhancement
- 2024Structured Pattern Expansion with Diffusion Models
- 2023MatFuse: Controllable Material Generation with Diffusion Models
- 2023Single Mesh Diffusion Models with Field Latents for Texture Generation
- 2025Channel-wise Noise Scheduled Diffusion for Inverse Rendering in Indoor Scenes
- 2023On the Noise Scheduling for Generating Plausible Designs with Diffusion Models
- 2024Denoising Monte Carlo Renders with Diffusion Models