Deep learning for image inpainting: A survey

作者：

Highlights：

• We summarize existing deep learning-based image inpainting algorithms in three aspects, including inpainting strategies, network structures and loss functions.

• We introduce the open source codes, popular used datasets, evaluation metrics, and application scenarios.

• We compare the inpainting algorithms with released source codes.

• We outline the challenges and possible future directions.

摘要

•We summarize existing deep learning-based image inpainting algorithms in three aspects, including inpainting strategies, network structures and loss functions.•We introduce the open source codes, popular used datasets, evaluation metrics, and application scenarios.•We compare the inpainting algorithms with released source codes.•We outline the challenges and possible future directions.

论文关键词：Image inpainting,Image restoration,Generative adversarial network,Convolutional neural network

论文评审过程：Received 22 December 2021, Revised 16 July 2022, Accepted 15 September 2022, Available online 20 September 2022, Version of Record 2 October 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.109046