@article{Lehmann_2022, title={Attention-based U-Net for image demoiréing}, volume={31}, url={https://mgv.sggw.edu.pl/article/view/4527}, DOI={10.22630/MGV.2022.31.1.1}, abstractNote={<p>Image demoiréing is a particular example of a picture restoration problem. Moiré is an interference pattern generated by overlaying similar but slightly offset templates.<br />In this paper, we present a <em>deep learning</em> based algorithm to reduce moiré disruptions. The proposed solution contains an explanation of the <em>cross-sampling</em> procedure – the training dataset management method which was optimized according to limited computing resources.<br />Suggested neural network architecture is based on <em>Attention U-Net</em> structure. It is an exceptionally effective model which was not proposed before in image demoiréing systems. The greatest improvement of this model in comparison to <em>U-Net</em> network is the implementation of <em>attention gates</em>. These additional computing operations make the algorithm more focused on target structures.<br />We also examined three MSE and SSIM based loss functions. The SSIM index is used to predict the perceived quality of digital images and videos. A similar approach was applied in various computer vision areas.<br />The author’s main contributions to the image demoiréing problem contain the use of the novel architecture for this task, innovative two-part loss function, and the untypical use of the <em>cross-sampling</em> training procedure.</p>}, number={1/4}, journal={Machine Graphics and Vision}, author={Lehmann, Tomasz}, year={2022}, month={Dec.}, pages={3–17} }