A Novel Dehazing Method for Color Fidelity and Contrast Enhancement on Mobile Devices

• Published in: IEEE transactions on consumer electronics Vol. 65; no. 1; pp. 47 - 56
• Main Authors: Wang, Ping, Fan, Qinglan, Zhang, Yunfeng, Bao, Fangxun, Zhang, Caiming
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.02.2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Atmospheric modeling; Color; Density; Digital cameras; Distortion; Electronic devices; Fog; Haze; Image color analysis; Image contrast; Image enhancement; Image restoration; Image segmentation; Image transmission; iterative algorithm; Iterative algorithms; Iterative methods; local atmospheric light; Meteorology; Mobile handsets; region division; Single image dehazing; Visibility; Weather
• ISSN: 0098-3063; 1558-4127
• DOI: 10.1109/TCE.2018.2884794

Mobile devices that can capture images outdoors often suffer from poor weather conditions. Observed scenes lose contrast and visibility, and are also subject to color shifts due to the presence of fog and haze. In this paper, a single image dehazing method for removing color distortion and enhancing contrast on mobile devices is presented. The method can effectively remove haze from images with an uneven haze density. First, based on the local haze density, a superpixel segmentation algorithm is employed to adaptively divide a hazy image into several regions. Second, the local atmospheric light in each image region is estimated, and an initial transmission map is calculated. Finally, based on the haze imaging model, an iterative algorithm is developed to estimate the local atmospheric light and transmission for restoring the image. The proposed method can effectively recover detailed image information, resulting in an image with natural color, with an acceptable time performance. Experimental results show that the proposed method can be applied effectively to mobile devices, including digital cameras and mobile phones for enhancing the contrast and visibility without introducing color distortion.