使用局部大气光的单幅图像去雾算法

Single Image Dehazing Algorithm Using Local Atmospheric Light

针对传统算法去雾后图像偏暗的问题,根据去雾后图像对比度和亮度均应该增加的标准提出了一种高亮度和对比度的去雾算法。首先依据大气光与复原图像亮度成反比事实,设置像素红绿蓝(RGB)三个颜色分量均值为局部级粗糙大气光,使用具有较好抑制光晕效应的半全局加权最小二乘算法优化获取局部级大气光;然后根据去雾图像方差与大气透射率成反比事实,使用基于像素最小颜色分量的大气散射函数计算粗糙透射率,然后使用局部均值滤波器优化透射率;最后结合雾天成像模型恢复无雾图像。实验结果表明所提算法去雾图像的梯度、亮度和速度等客观指标均优于传统算法,其中亮度最少增强1.5倍。所提算法能提高去雾图像亮度、丰富图像细节和提高去雾速度,具有较好的工程应用价值。

Aiming at the problem that hazy images will darken after dehazing,a dehazing algorithm with high brightness and contrast was proposed.First,based on the fact that atmospheric light was inversely proportional to the brightness of the restored image,the average value of the red,green,blue(RGB)color channels of the pixel was used as local raw atmospheric light,and the atmospheric light was optimized by the semi-global weighted least square algorithm.Then according to the fact that the dehazing image variance was inversely proportional to the atmospheric transmission map,and the atmospheric scattering function based on the minimum color channel of pixels was used to calculate the raw transmission map,and then used the local mean filter to optimize transmission map.Finally,the haze-free image was restored with the Koschmieder model.The experimental results show that the gradient,brightness and speed of the proposed algorithm are better than traditional algorithm,and the brightness is enhanced by at least 1.5 times.It has the advantages of high brightness,rich details and fast dehazing speed.