Efficient deep neural network for photo-realistic image super-resolution

作者：

Highlights：

• We develop a lightweight deep learning model for photo-realistic super-resolution.

• It requires 25% fewer operations than others while achieving better performance.

• Our method can handle multiple scale factors with a single network.

• We provide various analyses such as initialization schemes or efficiency trade-off.

摘要

•We develop a lightweight deep learning model for photo-realistic super-resolution.•It requires 25% fewer operations than others while achieving better performance.•Our method can handle multiple scale factors with a single network.•We provide various analyses such as initialization schemes or efficiency trade-off.

论文关键词：Super-resolution,Photo-realistic,Convolutional neural network,Efficient model,Adversarial learning,Multi-scale approach

论文评审过程：Received 8 August 2019, Revised 4 March 2022, Accepted 11 March 2022, Available online 14 March 2022, Version of Record 19 March 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108649