One Noise to Rule Them All: Learning a Unified Model of Spatially-Varying Noise Patterns
Arman Maesumi,Dylan Hu,Krishi Saripalli,Vladimir Kim,Matthew Fisher,Soeren Pirk,Daniel Ritchie
DOI: https://doi.org/10.1145/3658195
IF: 6.2
2024-07-19
ACM Transactions on Graphics
Abstract: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 plus 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. Open-sourced materials can be found at https://armanmaesumi.github.io/onenoise/
computer science, software engineering