Rethinking Noise Synthesis and Modeling in Raw Denoising

TL;DR

This paper presents a novel noise synthesis approach for raw image denoising that directly samples real sensor noise, improving realism and generalization over existing methods, and challenges recent DNN-based noise modeling conclusions.

Contribution

Introducing a sensor-specific noise sampling method with pattern-aligned patch sampling and high-bit reconstruction techniques for realistic raw noise synthesis.

Findings

Outperforms existing noise synthesis methods

Demonstrates wide generalization across sensors and lighting

Challenges effectiveness of DNN-based noise modeling methods

Abstract

The lack of large-scale real raw image denoising dataset gives rise to challenges on synthesizing realistic raw image noise for training denoising models. However, the real raw image noise is contributed by many noise sources and varies greatly among different sensors. Existing methods are unable to model all noise sources accurately, and building a noise model for each sensor is also laborious. In this paper, we introduce a new perspective to synthesize noise by directly sampling from the sensor's real noise. It inherently generates accurate raw image noise for different camera sensors. Two efficient and generic techniques: pattern-aligned patch sampling and high-bit reconstruction help accurate synthesis of spatial-correlated noise and high-bit noise respectively. We conduct systematic experiments on SIDD and ELD datasets. The results show that (1) our method outperforms existing methods and demonstrates wide generalization on different sensors and lighting conditions. (2) Recent conclusions derived from DNN-based noise modeling methods are actually based on inaccurate noise parameters. The DNN-based methods still cannot outperform physics-based statistical methods.