Multi-spectral Dynamic Feature Encoding Network for Image Demoiréing

Qiang Dai, Xi Cheng, Li Zhang

Abstract: Moiré often appears when photographing textured objects, which can seriously degrade the quality of the captured photos. Due to the wide distribution of moiré and the dynamic nature of the moiré textures, it is difficult to effectively remove the moiré patterns. In this paper, we propose a multi-spectral dynamic feature encoding (MSDFE) network for image demoiréing. To solve the issue of moiré with distributed frequency spectrum, we design a multi-spectral dynamic feature encoding module to dynamically encode moiré textures. To remedy the issue of moiré textures with dynamic nature, we utilize a multi-scale network structure to process moiré images at different spatial resolutions. Extensive experimental results indicate that our proposed MSDFE outperforms the state-of-the-art in terms of fidelity and perceptual on benchmarks.