Nighttime Dehazing Method Based on Adaptive Ambient Light Estimation Considering the Effects of Multiple Light Sources

In recent years, active research has been conducted on dehazing methods to remove fog and haze from images. Conventional dehazing methods can produce good results for images taken during the daytime. Existing methods face a significant challenge when estimating ambient light for nighttime images. In this study, we propose a nighttime dehazing method based on ambient light estimation that considers the characteristics of both light source and non-light source regions. The proposed method adaptively estimates ambient light by fusing the results with a weighted maximum filter and a guided filter based on a light source region weight map. Our experimental results demonstrate the proposed method’s superior performance.

I. Introduction

Images taken under adverse weather conditions, such as fog and haze, significantly reduce the visibility of the subject due to their effects. This adversely affects the performance of vision systems for video surveillance, remote sensing, and automated driving technology. To address these problems, methods for removing fog and haze in a single image (haze removal, dehazing) have attracted considerable interest in recent years.