中文摘要：

基于暗原色先验的去雾算法估计出的透射图清晰度低,采用软抠图优化透射图时算法复杂度高,耗时大。而应用经典引导滤波优化时,算法复杂度低,但恢复出的图像在景深跳变处存在光晕现象。为解决上述问题,提出一种快速去雾算法,将颜色衰减先验代替暗原色先验,通过建立线性模型简化求解透射图的过程,从而快速得到原始估计的透射图。将估计的透射图作为引导滤波的输入图像,先对其下采样再进行滤波处理,以降低算法复杂度。对输出图像的平坦区域和边缘区域采用不同的插值算法进行插值,得到原始大小的透射图。实验结果表明,该算法能够快速有效地恢复出无雾图像以及图像的细节信息和色彩,在保证图像质量的前提下,大幅提高算法的运行效率,降低算法复杂度,满足实际应用中的实时性要求。

英文摘要：

The transmission map estimated by the haze removal algorithm based on dark channel prior has low definition. The algorithm using soft matting to optimize the transmission map has high complexity and is time consuming. The complexity of the algorithm is low when the classical bootstrap filtering optimization is used ,but the recovered image has halos at jump in depth of field. In order to solve these problems,this paper proposes a fast haze removal algorithm. Color attenuation prior is used to replace dark channel priors. A linear model is established to simplify the process of solving transmission map, thus a transmission map of the original estimation is obtained quickly. The estimated transmission map is used as the input image of the guided filtering. Through down sampling and filtering, the complexity of the algorithm is reduced. Different interpolation algorithms are used to interpolate the flat region and the edge region of the output image, so as to get the transmission map with the original size. Experimental results show that the algorithm can restore the fog-free image as well as the details and color of the image fast and effectively. Under the premise of guaranteeing the image quality, the operation efficiency of the algorithm is greatly improved, and the algorithm complexity is reduced, which meets the real-time requirements in practical applications.