Summary
International Workshop on Smart Info-Media Systems in Asia
2021
Session Number:RS1
Session:
Number:RS1-1
Nighttime Single Image Haze Removal Employing Multi-Scale Gaussian Filter
Takafumi Akita, Seiichi Kojima, Eiji Uchino, Noriaki Suetake,
pp.18-21
Publication Date:2021/9/20
Online ISSN:2188-5079
DOI:10.34385/proc.66.RS1-1
PDF download (4.3MB)
Summary:
Recently, many studies on haze removal have been conducted. For a daytime hazy image, there are many effective dehazing algorithms. However, these methods can’t effectively remove haze for a nighttime hazy image. In nighttime conditions, there may be multiple light sources such as street lamps and neon signs. In such conditions, the atmospheric light tends to be non-uniform in the entire scene. Due to the non-uniform atmospheric light, nighttime haze removal is a very challenging problem. In this paper, we propose a nighttime haze removal method based on the Retinex theory. In the proposed method, the atmospheric light estimation is achieved based on the assumption that the atmospheric light varies smoothly in space by the analogy of the retinex theory. Concretely, the atmospheric light is estimated as the maximum value of outputs of multi-scale Gaussian filters. Through experiments with some images, we verify the effectiveness of the proposed method.