A hierarchical airlight estimation method for image fog removal. (August 2015)

• Record Type: Journal Article
• Authors: Cheng, Fan-Chieh; Cheng, Chung-Chih; Lin, Po-Hsiung; Huang, Shih-Chia
• Is Part Of: Engineering applications of artificial intelligence. Volume 43(2015:Jul.)
• Page Start: 27
• Page End: 34
• Publication Date: 2015-08
• Subjects: Defogging -- Physical optic model -- Airlight estimation; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2015.03.011
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: Fog phenomena result in airlight generation and degrade the visibility of the color image captured from the camera. To improve visibility, airlight estimation is necessary for image fog removal. As airlight is very bright, the traditional methods directly select bright pixels for airlight estimation. However, some bright pixels generated by light sources, such as train headlights, may interfere with the accuracy of the above-mentioned methods. In this paper, we propose a new airlight estimation method. Based on Gaussian distribution, the proposed method selects the airlight candidates in the brightest region of the input image. Moreover, the color similarity estimation is also applied to hierarchically refine the candidates. We then compute the average color from the refined candidate pixels for airlight estimation. Experimental results demonstrate that the proposed method is more accurate than other airlight estimation methods and has low time complexity.