颜色衰减结合暗通道先验的浓雾影像去雾算法

Dehazing Algorithm for Images with Dense Haze Based on Color Attenuation Prior and Dark Channel Prior

山区地表湿气在抬升作用和温度共同作用下容易形成雾气,会导致无人机在对地拍摄时获取到含有分布不均匀的浓雾的影像,本文提出了一种颜色衰减先验和暗通道先验相结合的方法,可以有效去除影像中的浓雾部分。首先基于颜色衰减先验理论,将亮度与饱和度之差作为雾浓度判别依据,借助最大类间方差法对影像中的浓雾与薄雾区域分割;其次根据分割结果采用不同的暗通道窗口和局部大气光计算暗通道值和透射率;最后对去雾后的影像进行亮度增强处理。实验结果表明,该方法能有效去除影像的浓雾部分,相比对照的暗通道法,信息熵、平均梯度、有效细节强度、峰值信噪比分别提升了1.17%、9.01%、11.31%、59.64%,对于无人机影像处理有较好的工程应用价值。

The surface moisture in the mountainous area is easy to become fog under the effects of uplift and temperature. That will cause the drone to obtain images with uneven distribution of dense fog. In order to solve this problem, this paper proposes a method combining color attenuation prior and dark channel prior. Firstly, based on the color attenuation prior theory, the difference between brightness and saturation is used as the basis for judging the haze density. Then the image is di-vided into regions using Ostu algorithm. Secondly, according to the segmentation results, different local atmospheric light and dark channel windows are used to calculate the dark channel value and transmittance. Finally, the brightness enhancement process is performed on the image. The results show that the method can effectively remove the foggy part of the image. Compared with the dark channel prior method, the information entropy, average gradient, effective detail intensity, and peak signal-to-noise ratio are increased by 1.17%, 9.01%, 11.31%, and 59.64%, respectively. It provides engineering value and applicability for drone image processing.