Abstract
Outdoor images are used in many domains, such as surveillance, geospatial mapping, and autonomous vehicles. The occurrence of noise in outdoor images is a widely observed phenomenon. They are primarily attributed to extreme natural and manufactured meteorological conditions, such as haze, smog, and fog. In autonomous vehicle navigation, recovering the ground truth image is essential, enabling the system to make more informed decisions. Accurate air-light and transmission map calculation is vital in recovering the ground truth image. An efficient approach for image dehazing that utilizes the mean channel prior (MCP) is presented in this paper to estimate the transmission map, followed by Gamma transformation to correct the transmission map obtained by MCP. This paper presents two novel contributions: first, an Alexnet network transfer model classification of hazy images as a preprocessing, and second, an efficient image dehazing based on an image fusion strategy. In the image dehazing stage, the transmission map estimated by the mean channel is altered with Gamma correction first. Then, the initial transmission map and its modified copy are combined using the weighted average fusion technique to retain the information in the initial transmission map. Additionally, the fused transmission map undergoes filtration using a guided filter to mitigate block and halo artifacts within the dehazed image. Lastly, the dehazed image is recovered using the improved transmission map by utilizing an optical scattering model. The proposed Alexnet network transfers algorithm significantly and decreases the quantity of training data required compared to the traditional classification algorithm. In addition, the network's classification accuracy can reach 98%. The proposed image dehazing showed better performance in terms of computational time, natural image quality evaluator (NIQE) index, peak-signal-to-noise ratio (PSNR), and structural similarity index (SSIM) than that of existing methods.
Recommended Citation
Al_airaji, Roa'a M.; ALRikabi, Haider TH. Salim; and Kamil, Rula
(2024)
"Classification and removal of hazy images based on a transmission fusion strategy using the Alexnet network,"
Karbala International Journal of Modern Science: Vol. 10
:
Iss.
2
, Article 14.
Available at:
https://doi.org/10.33640/2405-609X.3356
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.
