Salient object detection in low-light images via functional optimization-inspired feature polishing
作者：
Highlights：
• We propose an end-to-end network called LLISOD for salient object detection (SOD) in low-light images.
• We model an implicit nonlinear mapping from biased feature maps to salient objects.
• We design an unfolded implicit nonlinear mapping module for feature polishing.
• We contribute a new dataset covering 1502 images to explore SOD in low-light images.
摘要
•We propose an end-to-end network called LLISOD for salient object detection (SOD) in low-light images.•We model an implicit nonlinear mapping from biased feature maps to salient objects.•We design an unfolded implicit nonlinear mapping module for feature polishing.•We contribute a new dataset covering 1502 images to explore SOD in low-light images.
论文关键词：Low-light images,Salient object detection,Features polishing,Implicit nonlinear mapping,Functional optimization
论文评审过程：Received 2 June 2022, Revised 20 September 2022, Accepted 20 September 2022, Available online 24 September 2022, Version of Record 12 October 2022.
论文官网地址：https://doi.org/10.1016/j.knosys.2022.109938
