A priori-guided multi-layer rain-aware network for single image deraining

作者：

Highlights：

• A priori-guided rain-aware network (PGRAN) for single image deraining is proposed, which learns rain details in different directions by introducing guided Prewitt operator.

• A new detail network is designed for rain density-aware by concatenating multi-detail units, which learns rain details in different density levels.

• The peak signal to noise ratio and structural similarity based regularizers are introduced in the loss function to improve the deraining results.

• Numerical experiments are implemented on three datasets, two synthetic and a real rain dataset. Comparisons with several state-of-the-art methods are also tested and discussed.

摘要

•A priori-guided rain-aware network (PGRAN) for single image deraining is proposed, which learns rain details in different directions by introducing guided Prewitt operator.•A new detail network is designed for rain density-aware by concatenating multi-detail units, which learns rain details in different density levels.•The peak signal to noise ratio and structural similarity based regularizers are introduced in the loss function to improve the deraining results.•Numerical experiments are implemented on three datasets, two synthetic and a real rain dataset. Comparisons with several state-of-the-art methods are also tested and discussed.

论文关键词：Rain removal,Deep learning,Image restoration

论文评审过程：Received 19 June 2021, Revised 13 October 2021, Accepted 16 October 2021, Available online 26 October 2021, Version of Record 29 October 2021.

论文官网地址：https://doi.org/10.1016/j.knosys.2021.107613