Related Experiment Video
Updated: Aug 5, 2026
High-Throughput Total Internal Reflection Fluorescence and Direct Stochastic Optical Reconstruction Microscopy Using a Photonic Chip
Published on: November 16, 2019
Multi-Exposure HDR Imaging: A Review of Pixel-Level and Feature-Level Reconstruction Methods
Qian Tao1, Wei Wang1, Chaobing Zheng2
1Hubei Province Key Laboratory of Intelligent Information Processing and Real-Time Industrial System, School of Computer Science and Technology, Wuhan University of Science and Technology, Wuhan 430081, China.
This survey categorizes multi-exposure fusion (MEF) and ghost removal techniques for high-dynamic-range (HDR) imaging. It details deep learning methods, comparing pixel-space and feature-space approaches for artifact reduction in dynamic scenes.
Area of Science:
- Computer Vision
- Image Processing
- Computational Photography
Background:
- High-dynamic-range (HDR) imaging captures scenes with extreme lighting variations.
- Multi-exposure imaging is crucial for HDR, but temporal gaps cause ghosting artifacts in dynamic scenes.
Purpose of the Study:
- To systematically review and categorize existing literature on multi-exposure fusion (MEF) and ghost removal for HDR imaging.
- To provide a taxonomy of deep learning-based methods for ghost artifact mitigation.
Main Methods:
- Literature review and categorization of conventional and data-driven methods.
- Taxonomy of deep learning approaches based on alignment and fusion domains (pixel-space vs. feature-space).
- Comparison of representative works across different supervision settings.
Main Results:
- Categorization of methods into pixel-space (explicit motion compensation) and feature-space (implicit alignment) deep learning approaches.
- Summary of key design principles for effective ghost removal.
- Overview of common datasets and evaluation metrics for HDR imaging.
Conclusions:
- Deep learning offers advanced solutions for ghost artifact reduction in HDR imaging.
- Understanding the trade-offs between pixel-space and feature-space methods is crucial.
- Future research should address major bottlenecks and explore new directions in HDR imaging.
