Towards General Low-Light Raw Noise Synthesis and Modeling

TL;DR

This paper introduces a novel generative approach for modeling and synthesizing low-light raw noise, capturing complex sensor-specific noise characteristics across different ISO levels, and demonstrates improved denoising performance.

Contribution

It proposes a physics- and learning-based noise synthesis model that generalizes across sensors and ISO levels, along with a new discriminator and dataset for benchmarking.

Findings

Generated noise closely matches real noise distribution.

The method outperforms state-of-the-art denoising techniques.

Effective across various sensors and ISO settings.

Abstract

Modeling and synthesizing low-light raw noise is a fundamental problem for computational photography and image processing applications. Although most recent works have adopted physics-based models to synthesize noise, the signal-independent noise in low-light conditions is far more complicated and varies dramatically across camera sensors, which is beyond the description of these models. To address this issue, we introduce a new perspective to synthesize the signal-independent noise by a generative model. Specifically, we synthesize the signal-dependent and signal-independent noise in a physics- and learning-based manner, respectively. In this way, our method can be considered as a general model, that is, it can simultaneously learn different noise characteristics for different ISO levels and generalize to various sensors. Subsequently, we present an effective multi-scale discriminator termed Fourier transformer discriminator (FTD) to distinguish the noise distribution accurately. Additionally, we collect a new low-light raw denoising (LRD) dataset for training and benchmarking. Qualitative validation shows that the noise generated by our proposed noise model can be highly similar to the real noise in terms of distribution. Furthermore, extensive denoising experiments demonstrate that our method performs favorably against state-of-the-art methods on different sensors.