Doing More With Moiré Pattern Detection in Digital Photos

Area of Science:

• Computer Vision
• Digital Image Processing
• Machine Learning

Background:

• Moiré patterns degrade digital image quality.
• Accurate detection of moiré is crucial for image quality evaluation and artifact removal (demoiréing).

Purpose of the Study:

• To present a simple and efficient framework for extracting moiré edge maps from images containing moiré patterns.
• To develop a robust neural network for precise moiré pattern detection.

Main Methods:

• A novel strategy for generating training triplets: natural image, moiré layer, and synthetic mixture.
• Development of a Moiré Pattern Detection Neural Network (MoireDet) utilizing three encoders.
• Ensuring pixel-level alignment and accommodating diverse moiré pattern characteristics during training.

Main Results:

• MoireDet demonstrated superior identification precision of moiré images across two datasets.
• The framework significantly improved the performance of state-of-the-art demoiréing methods.
• Effective extraction of moiré edge maps was achieved.

Conclusions:

• The proposed framework and MoireDet offer an efficient solution for moiré pattern detection.
• The method provides valuable priors for image quality evaluation and demoiréing tasks.
• MoireDet's encoder design effectively leverages contextual and structural features for moiré analysis.