Nighttime image dehazing based on Retinex and dark channel prior using Taylor series expansion

Haze removal from nighttime images is more difficult compared with daytime image dehazing due to the uneven illumination, low contrast and severe color distortion. In this paper, following the approaches based on Dark channel prior, we propose a simple yet effective approach using Retinex theory and Taylor series expansion for nighttime image dehazing, referred to as ‘RDT’. Existing nighttime image dehazing methods do not handle color shift and glow removal very well. In order to address these issues, we first propose to decompose the atmospheric light image from the input image based on the Retinex theory. Taylor series expansion is then introduced for the first time to accurately estimate the pointwise transmission map. Finally, during the following processes of image fusion and color transfer, the atmospheric light image and potential haze-free image are adopted to obtain the final haze-free image. The experimental results on benchmark nighttime haze images demonstrate the superior performance of our proposed RDT dehazing method over the state-of-the-art methods.