Nighttime Image Dehazing based on Bright and Dark Channel Prior and Gaussian Mixture Model
Jiazhi Yu, Jun Zhang, Biyuan Li, Xiaochang Ni, Jianqiang Mei · 2023
A new method based on bright and dark channel prior and gaussian mixture model is presented for nighttime dehazing from a single image. Firstly, the nighttime image is divided into the sky area and other area using the Gaussian mixture model. At the same time, two different channel maps are obtained for the input image using the bright and dark channels. For the sky area, the prior transmittance of the bright channel is selected, and for other area, the prior transmittance of the dark channel is selected. The two are fused to obtain a new transmittance. Then, the restored image can be obtained from the atmospheric scattering model. At the same time, in order to enhance the brightness of the image. The final result is obtained by adaptive contrast enhancement of the brightness component of the restored image. The experimental results show that the proposed method is superior to several existing dehazing removal techniques in terms of luminous artifacts, color shift, overexposure and noise amplification. It can be effectively improved in dealing with similar problems such as blurring of photos taken by night traffic and outdoor monitoring.