Low-Light Image Enhancement Conditioned On Hierarchical Illumination Proposal
- Resource Type
- Conference
- Authors
- Hu, Shengjie; Li, Feng; Sun, Shiping; Chen, Hua; Yan, Wenbin; Zhang, Xiaogang
- Source
- 2023 China Automation Congress (CAC) Automation Congress (CAC), 2023 China. :3381-3386 Nov, 2023
- Subject
- Aerospace
Communication, Networking and Broadcast Technologies
Components, Circuits, Devices and Systems
Computing and Processing
Power, Energy and Industry Applications
Robotics and Control Systems
Signal Processing and Analysis
Transportation
Image resolution
Graphical models
Noise reduction
Lighting
Estimation
Generators
Proposals
low-light enhancement
Retinex model
encoder-decoder
- Language
- ISSN
- 2688-0938
Low-light enhancement is a challenging task that requires improving contrast, enhancing illumination, and resolving degradations simultaneously. We model the enhancement process following the simplified Retinex model and employ V-channel image to obtain illumination proposals of multilevels by utilizing the pixel-wise illumination enhancement curves (LE-curve). We then embed the illumination proposals features into the encoder-decoder architecture hierarchically by designed Pyramid illumination proposal embedding module (PIPEM) which contains a stacked spatial pyramid pooling (SSPP) module and four proposal embedding attention module (PEA) with different scales to align and interact cross-domain features. In addition, a novel fusion net is proposed to suppress noise problem. Extensive experiments demonstrate that the proposed method achieves competitive and impressive performance with existing advanced methods quantitatively and qualitatively.