Modeling Dual-Exposure Quad-Bayer Patterns for Joint Denoising and Deblurring

TL;DR

This work tackles simultaneous denoising and deblurring by leveraging a novel dual-exposure Quad-Bayer sensor and a dedicated restoration network, QRNet. It introduces a B2QB sampling method to generate aligned Quad-Bayer/RGB pairs and a Quad-Bayer to RGB pipeline with a frequency-aware loss, achieving state-of-the-art results on synthetic and real data. The proposed QRNet demonstrates superior restoration quality with lower computational cost and robustness across noise and motion blur scenarios, supported by a new QR dataset of Quad-Bayer–RGB pairs. The approach offers a practical RAW-space solution for high-fidelity imaging under challenging real-world conditions.

Abstract

Image degradation caused by noise and blur remains a persistent challenge in imaging systems, stemming from limitations in both hardware and methodology. Single-image solutions face an inherent tradeoff between noise reduction and motion blur. While short exposures can capture clear motion, they suffer from noise amplification. Long exposures reduce noise but introduce blur. Learning-based single-image enhancers tend to be over-smooth due to the limited information. Multi-image solutions using burst mode avoid this tradeoff by capturing more spatial-temporal information but often struggle with misalignment from camera/scene motion. To address these limitations, we propose a physical-model-based image restoration approach leveraging a novel dual-exposure Quad-Bayer pattern sensor. By capturing pairs of short and long exposures at the same starting point but with varying durations, this method integrates complementary noise-blur information within a single image. We further introduce a Quad-Bayer synthesis method (B2QB) to simulate sensor data from Bayer patterns to facilitate training. Based on this dual-exposure sensor model, we design a hierarchical convolutional neural network called QRNet to recover high-quality RGB images. The network incorporates input enhancement blocks and multi-level feature extraction to improve restoration quality. Experiments demonstrate superior performance over state-of-the-art deblurring and denoising methods on both synthetic and real-world datasets. The code, model, and datasets are publicly available at https://github.com/zhaoyuzhi/QRNet.

Figures (11)

• Figure 1: Illustration of (a) Normal Bayer pattern, (b-h) Other Bayer patterns with single exposure, (i) Quad-Bayer pattern with both long and short exposures, and (j) Image capturing process of dual-exposure Quad-Bayer images.
• Figure 2: Illustration of (a) Quad-Bayer pre-processing, (b) QRNet architecture (level 1 to 5 from top to bottom), and (c) input enhancement block.
• Figure 3: Illustration of Pixel Unshuffle and Pixel Shuffle operations shi2016real for Quad-Bayer images, which are invertible.
• Figure 4: Illustration of the data capturing workflow for a moving image and a static image from an edited video.
• Figure 5: Illustration of (a) Bayer to Quad-Bayer pattern sampling method (B2QB); (b) synthetic Quad-Bayer patterns, their corresponding RGB images, and four local patches extracted from the full resolution images. To better illustrate those images, we post-process them to have the same illuminations.