De2Net: Under-display camera image restoration with feature deconvolution and kernel decomposition

Hangyan Zhu, Shaohui Liu, Ming Liu, Zifei Yan, Wangmeng Zuo

Abstract: Highlights•Proposed a novel De2<math><msup is="true"><mrow is="true"></mrow><mrow is="true"><mn is="true">2</mn></mrow></msup></math>Net framework for UDC image restoration.•De2<math><msup is="true"><mrow is="true"></mrow><mrow is="true"><mn is="true">2</mn></mrow></msup></math>Net works no matter whether the PSFs are known or not.•Proposed a decomposed KPN to enlarge the receptive field while reducing memory usage.•De2<math><msup is="true"><mrow is="true"></mrow><mrow is="true"><mn is="true">2</mn></mrow></msup></math>Net performs favorably on three commonly used UDC image restoration datasets.