Related Experiment Video
Updated: Aug 16, 2026
14:09
High-Throughput Total Internal Reflection Fluorescence and Direct Stochastic Optical Reconstruction Microscopy Using a Photonic Chip
Published on: November 16, 2019
Deep image restoration in adverse weather: A survey
Zhenbo Song1, Ruixin Li1, Zhenyuan Zhang1
1Nanjing University of Science and Technology, Nanjing, China.
Summary
This survey comprehensively reviews deep learning methods for adverse weather image restoration. It covers single-task and All-in-One (AiO) models, datasets, and metrics, analyzing performance and future directions.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Image Processing
Background:
- Adverse weather conditions like haze, rain, and snow degrade image quality.
- Deep learning has advanced single-task image restoration but lacks generalization.
- All-in-One (AiO) models offer a unified approach for multiple weather degradations.
Purpose of the Study:
- To provide a comprehensive survey of deep learning-based adverse weather image restoration.
- To organize and review both single-task and AiO restoration models.
- To identify challenges and future research directions in the field.
Main Methods:
- Reviewing network architectures and learning paradigms for single-task and AiO models.
- Analyzing commonly used datasets, loss functions, and evaluation metrics.
- Summarizing benchmark results of representative methods on public datasets.
Main Results:
- Existing surveys lack a unified review of deep learning-based adverse weather restoration.
- This paper categorizes models by architecture and learning paradigms.
- Performance and generalization abilities of various methods are analyzed.
Conclusions:
- A comprehensive understanding of current deep learning approaches for adverse weather image restoration is presented.
- The survey highlights the need for improved generalization in restoration models.
- Key challenges and future research avenues are discussed to guide advancements.