Single Image Dehazing Algorithm Analysis with Hyperspectral Images in the Visible Range

Abstract

In foggy or hazy conditions, images are degraded due to the scattering and attenuation of atmospheric particles, reducing the contrast and visibility and changing the color. This degradation depends on the distance, the density of the atmospheric particles and the wavelength. We have tested and applied five single image dehazing algorithms, originally developed to work on RGB images and not requiring user interaction and/or prior knowledge about the images, on a spectral hazy image database in the visible range. We have made the evaluation using two strategies: the first is based on the analysis of eleven state-of-the-art metrics and the second is two psychophysical experiments with 126 subjects. Our results suggest that the higher the wavelength within the visible range is, the higher the quality of the dehazed images. The quality increases for low haze/fog levels. The choice of the best performing algorithm depends on the criterion prioritized by the metric design strategy. The psychophysical experiment results show that the level of agreement between observers and metrics depends on the criterion set for the observers’ task.

Visibility Enhancement and Fog Detection: Solutions Presented in Recent Scientific Papers with Potential for Application to Mobile Systems

“…Focusing on the fact that images are degraded in foggy or hazy conditions, the degradation depends on the distance, the density of the atmospheric particles, and the wavelength. The authors in [ 89 ] tested multiple single image dehazing algorithms and performed an evaluation based on two strategies: one based on the analysis of state-of-the-art metrics and the other one based on psychophysical experiments. The results of the study suggest that the higher the wavelength within the visible range, the higher the quality of the dehazed images.…”

“…Five well-known dehazing algorithms (dark channel prior (DCP) [ 31 ], Tarel method [ 18 ], Meng method [ 86 ], DehazeNet method [ 87 ], and Berman method [ 88 ]) were compared [ 89 ]. Two images under different levels of fog were used for testing alongside their corresponding fog-free original image (captured in the visible and near-infrared).…”

“…Medium haze conditions are considered. The table presents the ranking (from 1 to 5) Data from the table are extracted from[89].…”

Band Selection for Dehazing Algorithms Applied to Hyperspectral Images in the Visible Range

“…This degradation depends on the distance, the density of the atmospheric particles and the wavelength. We can differentiate between haze and fog by taking into account not only the radius size of water particles (less than 1 micron corresponds to haze) but also their concentration (more than 100 particles/cm 3 is considered as fog) [ 4 , 5 ]. In either case, these factors make it difficult to identify the main features of the objects recorded in the image, especially in distant scenes with high haze and/or fog density, which hinders further image-processing tasks due to poor visibility, whether performed by a human observer or by computer vision algorithms.…”

“…In the last decade, a large number of dehazing algorithms have been proposed, and this has become a growing area of research and development. Depending on the proposed paradigm, dehazing methods can be classified in different ways [ 5 , 20 , 21 ]. Thus, for example, it is usual to distinguish between methods based on image enhancement [ 22 , 23 , 24 , 25 , 26 ], methods based on image fusion [ 27 , 28 ], methods based on image restoration [ 29 ] and Deep-Learning based methods, which often combine different strategies in the network design [ 30 , 31 , 32 , 33 ].…”