MoiréNet: A Compact Dual-Domain Network for Image Demoiréing

2025-09-23

AI Summary

To address anisotropic, multi-scale moiré patterns arising from spectral aliasing between display pixel arrays and camera sensor grids during image acquisition, this paper proposes a compact dual-domain deep network. Methodologically, we design a directional frequency-spatial encoder to precisely identify moiré orientations and introduce a frequency-spatial adaptive selection mechanism for artifact suppression driven by learned features. The network adopts a lightweight U-Net architecture enhanced with directional differential convolution and dual-domain feature interaction modules. Evaluated on public benchmarks, our approach achieves state-of-the-art performance with only 5.513 million parameters—48% fewer than ESDNet-L—yielding significantly improved inference efficiency and deployment feasibility. The proposed method is particularly suitable for resource-constrained applications such as smartphone photography, industrial imaging, and augmented reality.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Mixture of Experts (MoE)
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Moiré patterns arise from spectral aliasing between display pixel lattices and camera sensor grids, manifesting as anisotropic, multi-scale artifacts that pose significant challenges for digital image demoiréing. We propose MoiréNet, a convolutional neural U-Net-based framework that synergistically integrates frequency and spatial domain features for effective artifact removal. MoiréNet introduces two key components: a Directional Frequency-Spatial Encoder (DFSE) that discerns moiré orientation via directional difference convolution, and a Frequency-Spatial Adaptive Selector (FSAS) that enables precise, feature-adaptive suppression. Extensive experiments demonstrate that MoiréNet achieves state-of-the-art performance on public and actively used datasets while being highly parameter-efficient. With only 5.513M parameters, representing a 48% reduction compared to ESDNet-L, MoiréNet combines superior restoration quality with parameter efficiency, making it well-suited for resource-constrained applications including smartphone photography, industrial imaging, and augmented reality.

Problem

• Removing anisotropic multi-scale artifacts from digital images
• Addressing spectral aliasing between display and camera sensors
• Developing parameter-efficient demoiréing for resource-constrained applications

Innovation

• U-Net-based dual-domain network for image demoiréing
• Directional Frequency-Spatial Encoder with difference convolution
• Frequency-Spatial Adaptive Selector for feature suppression