视频实时自适应去雾算法

摘要

Under climate condition of haze and fog, outdoor videos will show poor visibility, and real-time video dehazing is needed to restore the visibility of videos. Real-time video dehazing requires that single imagle haze removal algorithm has real-time speed, while existing methods are either not fast enough or have undesirable results. Furthermore, the extent of haze concentration on the scene captured in video will vary as time goes on. However, existing image dehazing algorithms use fixed parameters configured manually, which cannot make images under different degrees of haze concentration all have ideal results. This paper presents a real-time adaptive video dehazing algorithm. This algorithm designs an image dehazing method that distinguishes different regions of input image through dark channel value. It uses different parameters in different regions and achieves great results in real time. Besides, this algorithm designs a method to evaluate dehazing results based on dark channel prior and uses an iteration method to adjust parameters automatically to deal with the problem that haze in videos keeps changing.%在雾天环境下，户外视频的可视性将受到极大损害，需要通过视频实时去雾来恢复视频的可视性。视频实时去雾对于单帧图像处理的速度有很高的要求，现有的图像去雾算法或是速度上达不到要求，或是速度虽快但去雾效果不理想。另外，视频还会面临拍摄场景中雾气浓度不断变化的问题，现有图像去雾算法中需要手动设置参数且参数固定，无法在雾气浓度变化的条件下始终达到理想的去雾效果。提出了一种实时的视频自适应去雾算法，该算法对视频中单帧图像进行去雾时，会基于暗原色值来区分图像区域，并对不同区域进行不同程度的去雾，在满足实时性的同时得到了很好的去雾效果。此外，该算法还基于暗通道先验设计了评价去雾结果的方法，并使用迭代的方式根据雾气浓度自动调整去雾参数，从而在视频中雾气浓度变化的情况下，始终能达到理想的去雾效果。