Feature learning and patch matching for diverse image inpainting
作者：
Highlights：
• An exemplar-based inpainting is introduced to inpaint free-form holes with diverse plausible contents.
• Design a U-Net-like learning model for semantic inpainting with better color consistency.
• Apply patch matching on multiple exemplars to improve inpainting diversity and quality.
• Experiments demonstrate our proposed approach is effective for diverse image completion.
摘要
•An exemplar-based inpainting is introduced to inpaint free-form holes with diverse plausible contents.•Design a U-Net-like learning model for semantic inpainting with better color consistency.•Apply patch matching on multiple exemplars to improve inpainting diversity and quality.•Experiments demonstrate our proposed approach is effective for diverse image completion.
论文关键词：Diverse image inpainting,Free-form mask,U-Net-like network,Nearest neighbors
论文评审过程：Received 28 September 2020, Revised 30 March 2021, Accepted 10 May 2021, Available online 29 May 2021, Version of Record 13 June 2021.
论文官网地址：https://doi.org/10.1016/j.patcog.2021.108036
