Turkish Journal of Electrical Engineering and Computer Sciences
Outdoor captured scenes are degraded by atmospheric particles and water droplets. Due to scattering andabsorption effects in the atmosphere, degraded images lose contrast and color fidelity. Performances of the computervision algorithms are bound to suffer from low-contrast scene radiance. In many single-image dehazing models, thelarger the deviation in estimation of the key parameters such as transmission map and atmospheric light, the higherthe halo artifacts and loss of fine details in the dehazed image. The available models assume that the scattering lightis independent of wavelength, as the size of the atmospheric particles is larger compared to the wavelength of light.The model presented in this paper emphasizes the appropriate estimation of intensified transmission map from thehazy images by exploiting the scattering coe?icient in order to address the issues of haze concentrations. Experimentsconducted on thick and thin hazy images provide an optimal estimation of the model parameters, which can be applieddirectly in real-time situations. The available models are observed to be inconsistent sometimes in the enhancement ofcontrast, saturation and color information either together or independently. The proposed model addresses these issuesby extracting the haze-relevant features from the hazy images, such as hue disparity, contrast, and darkness, which yieldmore vivid saturation results. Moreover, the proposed model addresses different haze densities in the scene without theuse of refinement filters.
Dense haze, scattering coefficient of atmosphere, hue disparity, haze-level prior approach
THIRUMALA, VIJAYA LAKSHMI; KARANAM, VENKATA SATYANARAYANA; LANKIREDDY, PRATAP REDDY; KAKUMANI, ARUNA KUMARI; and YACHARAM, RAKESH KUMAR
"Haze-level prior approach to enhance object visibility under atmosphericdegradation,"
Turkish Journal of Electrical Engineering and Computer Sciences: Vol. 29:
2, Article 33.
Available at: https://journals.tubitak.gov.tr/elektrik/vol29/iss2/33
Computer Engineering Commons, Computer Sciences Commons, Electrical and Computer Engineering Commons