HazeSpace2M: A Dataset for Haze Aware Single Image Dehazing

Md Tanvir Islam,Nasir Rahim,Saeed Anwar,Muhammad Saqib,Sambit Bakshi,Khan Muhammad

DOI: https://doi.org/10.1145/3664647.3681382

2024-09-26

Abstract:Reducing the atmospheric haze and enhancing image clarity is crucial for computer vision applications. The lack of real-life hazy ground truth images necessitates synthetic datasets, which often lack diverse haze types, impeding effective haze type classification and dehazing algorithm selection. This research introduces the HazeSpace2M dataset, a collection of over 2 million images designed to enhance dehazing through haze type classification. HazeSpace2M includes diverse scenes with 10 haze intensity levels, featuring Fog, Cloud, and Environmental Haze (EH). Using the dataset, we introduce a technique of haze type classification followed by specialized dehazers to clear hazy images. Unlike conventional methods, our approach classifies haze types before applying type-specific dehazing, improving clarity in real-life hazy images. Benchmarking with state-of-the-art (SOTA) models, ResNet50 and AlexNet achieve 92.75\% and 92.50\% accuracy, respectively, against existing synthetic datasets. However, these models achieve only 80% and 70% accuracy, respectively, against our Real Hazy Testset (RHT), highlighting the challenging nature of our HazeSpace2M dataset. Additional experiments show that haze type classification followed by specialized dehazing improves results by 2.41% in PSNR, 17.14% in SSIM, and 10.2\% in MSE over general dehazers. Moreover, when testing with SOTA dehazing models, we found that applying our proposed framework significantly improves their performance. These results underscore the significance of HazeSpace2M and our proposed framework in addressing atmospheric haze in multimedia processing. Complete code and dataset is available on \href{<a class="link-external link-https" href="https://github.com/tanvirnwu/HazeSpace2M" rel="external noopener nofollow">this https URL</a>} {\textcolor{blue}{\textbf{GitHub}}}.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problems that this paper attempts to solve are some key limitations in existing data sets and methods in the field of single - image de - hazing. Specifically:

1. **Lack of foggy images in real environments**: Existing synthetic data sets often lack diverse types of fog, which makes de - hazing algorithms ineffective when dealing with different types of fog in actual scenes.
2. **Single - type fog data sets**: Most existing data sets contain only one type of fog, such as fog, cloud, or environmental fog, which limits the generalization ability of de - hazing algorithms.
3. **Lack of de - hazing methods specifically for different fog types**: Current methods usually do not distinguish between fog types and directly apply general - purpose de - hazing algorithms, which may lead to unsatisfactory de - hazing results.

To address these problems, the paper proposes the **HazeSpace2M** data set, which is a data set containing more than 2 million images, covering multiple scenes (outdoor, street, farmland, satellite) and three different fog types (fog, environmental fog, cloud), with 10 different fog intensity levels for each type. In addition, the paper also proposes an intelligent de - hazing framework based on fog type classification. By first identifying the fog type in the input image and then selecting the corresponding de - hazing algorithm, the de - hazing effect is improved.

### Specific objectives:
1. **Develop a comprehensive benchmark data set**: The HazeSpace2M data set aims to provide diverse scenes and fog types to support more effective fog type classification and de - hazing algorithm training.
2. **Propose an intelligent de - hazing framework**: This framework improves the de - hazing effect by identifying the fog type in the image and then applying a specific de - hazing algorithm.
3. **Evaluate the performance of existing models**: Use the HazeSpace2M data set to benchmark existing classification and de - hazing models and verify the effectiveness of the proposed framework.
4. **Demonstrate the advantages of specialized de - hazing algorithms**: Experimental results show that specialized de - hazing algorithms based on accurate fog type classification can significantly improve the de - hazing effect, surpassing general - purpose de - hazing algorithms.

### Main contributions:
1. **Developed a comprehensive benchmark data set**: HazeSpace2M contains more than 2 million images, covering multiple scenes and fog types, and is one of the largest foggy image data sets currently.
2. **Proposed an intelligent de - hazing framework**: This framework improves the de - hazing effect through the combination of fog type classification and specialized de - hazing algorithms.
3. **Provided new benchmark test results**: Evaluated existing classification and de - hazing models, showing the advantages of the HazeSpace2M data set and the proposed framework.
4. **Verified the effectiveness of specialized de - hazing algorithms**: Experimental results show that specialized de - hazing algorithms based on accurate fog type classification perform better in terms of metrics such as PSNR, SSIM, and MSE.

These contributions not only promote research in the field of single - image de - hazing, but also provide new solutions to the atmospheric fog problem in multimedia processing.