Low-Light Image Dehazing with Variational Enhancement Model
Chao Yan, Yin Gao, Hao Li, Jun Li · 2023
Low-light image dehazing is a challenging and demanding task. Existing nighttime dehazing methods often suffer from issues like halo artifacts, color shifts, and amplified noise. To obtain natural haze-free images, this paper proposes a simple yet effective dehazing method based on a variational enhancement model. First, we reduce the influence of artificial light sources commonly found in low-light hazy images through glow correction. Then, the variational model based on cosine-gaussian is constructed to enhance the brightness of low-light hazy images and make the details more prominent. Finally, various factors such as the dark channel, channel difference, and brightness features are considered to solve the atmospheric scattering model and obtain refined haze-free images. Quantitative and qualitative comparisons validate the superior performance of the proposed low-light dehazing method compared to state-of-the-art approaches. Furthermore, the proposed method also demonstrates significant effectiveness in daytime image dehazing.