IDACC: Image Dehazing Avoiding Color Cast Using a Novel Atmospheric Scattering Model

Abstract

Images captured in hazy scenes exhibit severe degradation. Various dehazing methods have been proposed recently. However, most of them have the drawback of color cast in the dehazing results. To solve this problem, we proposed an image dehazing algorithm called IDACC that avoids color cast by using a novel atmospheric scattering model (NASM). The NASM is obtained by introducing a scattering compensation coefficient into the atmospheric light term of the traditional atmospheric scattering model (ASM). The scattering compensation coefficient effectively suppresses the color cast caused by the atmospheric light term. Moreover, we designed a global information search (GIS) strategy based on the minimum value channel to facilitate the calculation of the transmission map. Extensive experiments show that IDACC achieves image dehazing and effectively avoids color cast. Note that the IDACC algorithm does not require to consume time in dataset collection and model training. Furthermore, IDACC needs to only input a hazy image and can directly output a haze-free image. Besides, our method shows excellent performance both quantitatively and qualitatively compared with state-of-the-art (SOTA) methods.

Keywords

• Image dehazing
• color cast
• scattering compensation coefficient
• atmospheric scattering model