A New Multi-Level Hazy Image and Video Dataset for Benchmark of Dehazing Methods

TL;DR

This paper introduces a new multi-level hazy image and video dataset to benchmark dehazing methods, highlighting the variability in performance between traditional and deep learning approaches across different haze levels.

Contribution

The study provides a controlled, multi-level hazy dataset with ground truth, enabling comprehensive benchmarking of various dehazing methods and revealing insights into their generalization capabilities.

Findings

Traditional dehazing methods outperform deep learning models on this dataset.

Deep models' performance varies significantly depending on the scene.

Cross-dataset generalization of deep models is generally poor.

Abstract

The changing level of haze is one of the main factors which affects the success of the proposed dehazing methods. However, there is a lack of controlled multi-level hazy dataset in the literature. Therefore, in this study, a new multi-level hazy color image dataset is presented. Color video data is captured for two real scenes with a controlled level of haze. The distance of the scene objects from the camera, haze level, and ground truth (clear image) are available so that different dehazing methods and models can be benchmarked. In this study, the dehazing performance of five different dehazing methods/models is compared on the dataset based on SSIM, PSNR, VSI and DISTS image quality metrics. Results show that traditional methods can generalize the dehazing problem better than many deep learning based methods. The performance of deep models depends mostly on the scene and is generally poor on cross-dataset dehazing.