Improved algorithm on haze removal based on dark channel prior and histogram specification
Dong Gang, Zhenyu Li, Wei Xu, Bo Yang, Sun Xiaobin, Guo Rui · 2018
In this paper, an improved algorithm of dark channel prior is proposed. In the bright areas such as the sky, the estimated transmittance of dark channel prior is low, so the processed image appears color distortion. Besides, the intensity of the recovered haze-free image through dark channel prior is dark and the image's contrast is low. The improved algorithm is proposed aimed at the two defects. Firstly, the transmittance is optimized through the guided filter. Then, the transmission of the bright areas is corrected by using the improved function. Finally, the intensity and contrast of the recovered image are adjusted through the Gaussian Histogram Specification. The experimental results show that the two defects of dark channel prior can be overcame by the improved algorithm of haze removal which can avoid the color distortion and enhance the visual effect of image.