Improving Complex Moiré Removal with Generative Supervision

Abstract

The availability of high-quality paired data is essential for training learning-based image demoiréing models. However, it remains challenging for existing datasets to encompass the complex moiré patterns captured in uncontrolled real-world scenarios. Such degradations typically manifest as large-scale, multicolored moiré patterns. Moreover, these patterns frequently occur in images for which clean counterparts are difficult to obtain, such as photographs acquired from public displays or existing online resources.

We propose a data engine designed to improve the removal of complex moiré patterns by generating training supervision. We first collect real-world images containing complex moiré patterns and localize the corresponding screen regions. Multiple image-conditioned generative foundation models then produce candidate references, which are subjected to patch-level quality control to filter and select reliable supervision. Based on this paradigm, we construct WildMoiré with 6.8K moiré–GT training pairs and an independent captured test set of approximately 250 pairs. Extensive experiments on ESDNet, SDXL, and Qwen-Image-Edit show consistent improvements in complex moiré removal.

Citation

If you find this work useful, please cite:

@misc{gu2026improving,
  title  = {Improving Complex Moiré Removal with Generative Supervision},
  author = {Xinyang Gu and Zhilu Zhang and Honglei Xu and Yanting Mei and Yukang Ding and Wangmeng Zuo},
  year   = {2026}
}