Traffic Image Dehazing Using Sky Segmentation and Color Space Conversion

• Published in: Journal of Information Science and Engineering Vol. 38; no. 1; pp. 223 - 251
• Main Authors: 郭璠(FAN GUO), 邱俊峰(JUN-FENG QIU), 唐琎(JIN TANG), 李伟清(WEI-QING LI)
• Format: Journal Article
• Language: Chinese; English
• Published: Taipei 社團法人中華民國計算語言學學會 01.01.2022; Institute of Information Science, Academia Sinica
• Subjects: Algorithms; Color; Conversion; Equalization; Haze; Histograms; Image contrast; Image filters; Image restoration; Image segmentation; image dehazing; traffic scene; sky segmentation; color space conversion; traffic applications
• ISSN: 1016-2364
• DOI: 10.6688/JISE.202201_38(1).0012

In order to restore degraded traffic images in haze and dark environment, we present an efficient traffic image haze removal method using sky segmentation and color space conversion. The dark channel++ and contrast energy++ features are proposed for the fast sky segmentation step. The atmospheric light is estimated based on the haze density in different region, and the dehazing procedure is executed in HSI color space. Besides, this method takes advantage of the contrast limited adaptive histogram equalization (CLAHE) and guided image filtering to ensure a visual pleasing result. The experimental results for both synthetic and natural hazy images demonstrate that our algorithm performs comparable or even better results than the state-of-the-art methods in terms of various measurement indexes, such as the MSE, SSIM, mean gradient change rate, etc. Two traffic applications, such as road-marking extraction and vehicle detection, are presented to verify the effectiveness of the proposed algorithm.