Image dehazing using adaptive bi-channel priors on superpixels

• Published in: Computer vision and image understanding Vol. 165; pp. 17 - 32
• Main Authors: Jiang, Yutong, Sun, Changming, Zhao, Yu, Yang, Li
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.12.2017
• Subjects: Bright channel prior; Dark channel prior; Dehazing; Image restoration; Superpixels; Bright channel prior; Superpixels; Image restoration; Dark channel prior; Dehazing
• ISSN: 1077-3142; 1090-235X
• DOI: 10.1016/j.cviu.2017.10.014

•We propose an adaptive bi-channel priors (ABiCP) method for image dehazing.•We propose a method for determining the white and black pixels in haze images based on the HSV space.•We introduce the superpixels in image dehazing to reduce computational time and improve the quality of restored images. Recently, a number of image dehazing methods are developed based on dark channel prior which is simple yet effective. In order to compensate for any failure on the use of dark channel prior in white regions and bright channel prior in black regions, an image dehazing method using a novel adaptive bi-channel priors on superpixels is presented in this paper. In the proposed method, a haze image is converted to the hue, saturation, and value space, and the linearly transformed thresholds on saturation and value are used to detect any white and black pixels. Using superpixels as local regions, the local transmission and atmospheric light values are estimated more reliably and efficiently by combining the dark and bright channel priors (bi-channel priors). Furthermore, adaptive bi-channel priors are developed to rectify any incorrect estimations on transmission and atmospheric light values for white and black pixels that fail to satisfy the assumptions of the bi-channel priors. After applying our dehazing method, the white and black pixels on the restored image are with excellent fidelity. Experimental results demonstrate that our proposed method performs better for restoring images in terms of both quality and execution speed than the current state-of-the-art dehazing methods.