基于自适应暗原色的单幅图像去雾算法

Single Image Dehazing Algorithm Based on Adaptive Dark Channel Prior

为了复原雾天退化图像,提出了一种自适应暗原色的单幅图像去雾算法.针对暗原色先验理论在估计图像透射率时不够准确、容易引起Halo效应的问题,采用自适应暗原色概念,即在暗原色的获取过程中引入自适应阈值,减小景深变化对暗原色获取的影响,进而正确求取透射率.此过程不需导向滤波的细化,也就避免了导向滤波引起的效率低和去雾不彻底的问题.主观及客观两方面将本文去雾算法与现有算法进行对比,结果表明,本文算法能够有效消除Halo效应,获得高对比度、高色彩饱和度以及丰富细节信息的去雾结果,同时也提高了图像去雾效率.

In order to recover the degraded image induced by the fog or haze, this paper proposes a single image dehazing algorithm based on adaptive dark channel prior. The error during the estimation of transmittance by Dark Channel Prior（DCP） will directly cause Halo effect. To deal with this problem, the notion of Adaptive Dark Channel Prior（ADCP） was proposed, it means using adaptive in the acquisition of DCP, it can reduce the effect brought by the change of depth of focus, So it will obtain the transmittance correctly without the use of Guided Filtering（GF）, this means it will avoid low efficiency and defog incomplete caused by the filtering. Experiments show that the improved dehazing algorithm could eliminate the Halo effect and achieve the dehazing image with high contrast, high color saturation and abundant details from both objective or subjective image-quality assessment. Meanwhile, the speed of image process is also improved.