Related Experiment Video
Updated: Aug 6, 2026
10:16
Digital Inline Holographic Microscopy (DIHM) of Weakly-scattering Subjects
Published on: February 8, 2014
Enhancing low-light images with MSHCDI-Net: A multi-scale hybrid cross-domain interaction approach
Bin Chen1, Peitao Li1, Chaobing Zheng1
1School of Electronic Information/Wuhan University of Science and Technology, Wuhan, Hubei, China.
Plos One
|July 17, 2026
Summary
This study introduces MSHCDI-Net, a novel network for low-light image enhancement. It effectively combines convolutional neural networks and Transformers to improve visibility and detail in dark images.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Image Processing
Background:
- Low-light image enhancement is crucial for visual quality.
- Conventional CNNs struggle with long-range dependencies, limiting global context.
- Restricted receptive fields in CNNs lead to suboptimal restoration.
Purpose of the Study:
- To develop a network that effectively models both local details and global context for low-light images.
- To overcome the limitations of CNNs in capturing long-range dependencies.
- To improve structural consistency and detail restoration in challenging illumination.
Main Methods:
- Proposed MSHCDI-Net (Multi-Scale Hybrid Cross-Domain Interaction Network).
- Integrated CNN and Transformer branches for joint feature extraction.
- Employed a hierarchical encoder-decoder architecture with cross-domain interaction.
- Utilized adaptive feature fusion and multi-scale guidance.
Main Results:
- MSHCDI-Net achieved significant improvements in low-light image enhancement.
- Demonstrated competitive performance on public benchmarks like LOL-v1 and LOL-v2.
- Achieved 23.45 dB PSNR / 0.848 SSIM on LOL-v1.
- Achieved 23.74 dB PSNR / 0.910 SSIM on LOL-v2-synthetic.
- Achieved 22.24 dB PSNR / 0.868 SSIM on LOL-v2-real.
Conclusions:
- MSHCDI-Net effectively captures local textures and global context.
- The hybrid approach enhances structural consistency and detail restoration.
- The proposed method offers superior quantitative and visual results for low-light enhancement.
