GuidNoise: Single-Pair Guided Diffusion for Generalized Noise Synthesis

TL;DR

GuidNoise addresses the challenge of realistic real-world noise synthesis without camera metadata by learning a single-pair guided diffusion. It introduces a cascade decoding architecture with Guidance-aware Affine Feature Modification (GAFM) and a noise-aware refine loss to capture and transfer noise distributions from a single guidance pair. The method achieves generalized noise synthesis across diverse datasets and enables effective self-augmentation to boost denoising performance for lightweight models and limited data. Experiments show GuidNoise yields noise distributions closer to real data (AKLD/KLD) and denoising performance approaching that of models trained on real noise.

Abstract

Recent image denoising methods have leveraged generative modeling for real noise synthesis to address the costly acquisition of real-world noisy data. However, these generative models typically require camera metadata and extensive target-specific noisy-clean image pairs, often showing limited generalization between settings. In this paper, to mitigate the prerequisites, we propose a Single-Pair Guided Diffusion for generalized noise synthesis GuidNoise, which uses a single noisy/clean pair as the guidance, often easily obtained by itself within a training set. To train GuidNoise, which generates synthetic noisy images from the guidance, we introduce a guidance-aware affine feature modification (GAFM) and a noise-aware refine loss to leverage the inherent potential of diffusion models. This loss function refines the diffusion model's backward process, making the model more adept at generating realistic noise distributions. The GuidNoise synthesizes high-quality noisy images under diverse noise environments without additional metadata during both training and inference. Additionally, GuidNoise enables the efficient generation of noisy-clean image pairs at inference time, making synthetic noise readily applicable for augmenting training data. This self-augmentation significantly improves denoising performance, especially in practical scenarios with lightweight models and limited training data. The code is available at https://github.com/chjinny/GuidNoise.

Figures (11)

• Figure 1: Illustration of GuidNoise. Given an input clean image and a noisy-clean guidance image pair, GuidNoise can generate a synthesized noisy image that mimics the real noisy image by capturing the noise distribution of the guidance image. Since the given images can be easily obtained from diverse environments, pseudo-real noisy images can be synthesized arbitrarily.
• Figure 1: Architecture overview of GuidNoise.
• Figure 2: Overview of the proposed method. The figure shows the training and inference pipelines of GuidNoise. The generation model $\eta(\cdot)$ synthesizes noisy images using decoder $\mathcal{D}$, which takes latent feature $\mathbf{Z_x}$ as input. Each residual block processes a concatenation of the previous decoder output and two encoder features, modulated by affine parameters $(\alpha_i, \beta_i)$.
• Figure 2: Architecture of Down/Up Blocks with skip connection.
• Figure 3: Qualitative comparison of synthetic noisy imagess on SIDD-Validation (SIDD), SIDD+, PolyU and Nam dataset.