Adaptive detail extraction network for Moiré pattern removal in screen photography

In this paper, we propose a neural-network-based demoiréing algorithm aimed at eliminating moiré patterns and enhancing image quality. Our algorithm designs an adaptive detail extraction block to capture fine texture information in images, and builds a composite loss function including the L1 loss, the perceptual loss, and the contrastive loss to enhance the network’s perceptual and representational capabilities across different scales. Experimental results on multiple datasets demonstrate that the proposed algorithm exhibits comparable or even superior demoiréing performance compared with state-of-the-art (SOTA) algorithms.