Nighttime image dehazing with a new light segmentation method and a linear image depth estimation model

Image with the scene of haze at night has low contrast, non-uniform illumination, color cast and significant noise. These cause nighttime haze removal from single image to be problematic and challenging. In this paper, we put forward a method that can remove nighttime haze from images and improve image quality. The input image is first decomposed into a glow layer and a haze layer with a modified color channel transformation for glow artifacts and color correction. A new light segmentation function is proposed next by using gamma correction of the channel difference and setting the threshold levels as the probability of a pixel belonging to a light source region. Then we estimate the ambient illuminance map by combining the maximum reflectance prior value with the aforementioned probability and computing the atmospheric light in the light and non-light regions. Finally, we establish a novel linear model to build the connection between the image depth map and three image features including luminance, saturation and gradient map for the light source regions while using the dark channel prior for the non-light source regions. The result of the light segmentation is 0.07, and the parameters of the linear depth estimation are 1.026 7, -0.596 6, 0.673 5 and 0.004 135. Experimental results show the proposed method is reliable for removing nighttime haze and glow of active light sources, reducing significant noise and improving visibility.