在单幅图像的条件下,为了能更快地实现雾天图像的复原, 且使其在雾天和霾天的状况下均能获得较好的复原效果,对基于暗原色先验的雾图还原方法和基于 人眼视觉理论的Retinex方法进行了结合和改进。对前者的大气光强度的估算进行改进,并对灰霾 天气状况下图像色彩进行矫正,使其对雾天和灰霾天气图像均适用。此外将暗原色先验理论中的透射率 估算进行简化,结合Retinex算法实现了雾霾图像的实时处理。经实验验证,该方法对道路监控等 应用场景下的雾霾图像处理有较好的效果,同时保证了处理的实时性。

In single image, in order to recover the image in the foggy weather faster, and make it suitable for both foggy and haze weather, the method of dark channel prior and the Retinex method based on the theory of human vision are combined and improved. Estimation of atmospheric light intensity in the dark channel prior is improved and the color of haze image is rectified in the method, which make it suitable for both the fog and haze image. In addition, the transmissivity estimation in the method of dark channel prior is simplified. The simplified dark channel prior method and the Retinex algorithm are combined, which achieves real-time processing of the fog and haze image. Based on the experiment, this method has preferable efficiency on processing the fog and haze image in the application of monitoring road, while the real-time speed is also guaranteed.