DGCA: high resolution image inpainting via DR-GAN and contextual attention

The most image inpainting algorithms often have existed problems such as blurred image, texture distortion and semantic inaccuracy, and the image inpainting effect is limited for images with large missing regions and resolution level. To solve above problems, the paper proposes an improved two-stage image inpainting network based on parallel network and contextual attention. Firstly, the improved deep residual network is used to perform generative pixels filling on the missing area, and the first-stage adversarial network is used to complete the edges information. Then, the color features of the filling map are extracted, the edge map is fused and complemented, and the fusion map is used as the conditional label of the second-stage adversarial network. Finally, the image repairing result has obtained through the two-stage network with the contextual attention module. The experiments on public datasets can show that the proposed algorithm can obtain a more realistic repairing effect. © 2023, The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.