基于引导滤波和暗原色先验理论透射率估值的视频去雾算法

针对雾天条件下户外采集的图像严重退化问题，解决传统的暗原色先验理论算法出现的边缘残雾、天空区域去雾效果欠佳、实时性差和鲁棒性差等问题，提出去雾效果显著的实时视频去雾算法.对大气光散射模型进行改进，以引导滤波后的灰度图作为大气光估计图；利用四叉树法和暗原色先验理论（DCP）在暗原色图中寻找浓雾区域，求得透射率估计值；利用改进的大气光散射模型复原图像.通过大量实验表明，复原出的图像去雾效果彻底，色彩鲜艳亮丽，天空区域不会出现彩色失真，景深变化大的地方不会出现白边现象，对于不同浓度的雾都有着较好的去雾效果，处理速度快且稳定，适合于实时视频去雾.

Real-time video dehazing using guided filtering and transmissivity estimated based on dark channel prior theory

An effective real-time video dehazing algorithm was proposed in view of serious degradation of outdoor images taken in foggy weather in order to overcome traditional dark channel prior algorithm's problems of remnant fog in edge, poor effect in the sky area and poor real-time and robustness. The algorithm modified the atmospheric light scattering model, and took gray-scale image after guided filtering as atmospheric light map. Then transmittance values of dense fog area were estimated by using the subdivision and dark channel prior theory algorithm on dark channel map. The image was restored by modified atmospheric light scattering model. Numerous experimental results show that the foggy image is clear and bright, and it won't appear any color distortion in sky area and white border in where depth of field change fast. A great dehazing effect can be achieved in pictures taken in different concentrations of fog. The algorithm is fast and stable, which is suitable for real-time video dehazing.