An overview of Image Dehazing Algorithms

In this paper, we present a broad study of all the state-of-the-art algorithms that have been published in the realm of image dehazing using deep learning to de-haze synthetic and real-world images. We further proceed to evaluate all of these algorithms based on various parameters as taken from all the papers and the data that has been reported in those papers. This paper aims to compare all these algorithms and also to shed light on the limitations of all the algorithms that have been included in the scope of this study. The papers that have been included in the scope of this review range from the early 2000s to as recently as 2021. The methods in each paper vary from the usage of Generative Adversative Networks to Zero-shot imaging and using Convolutional Neural Networks to taking the traditional approach of using dark channel priors to improve the results obtained. We have tabulated the results that have been recorded in all the papers included in this study. The papers have been evaluated on the basis of common image parameters such as Peak Signal to Noise Ratio (PSNR) and Structural Similarity Index (SSIM). At the end we look at the algorithms that perform the best in all these areas individually as well as the over-all best-performing algorithm.