一种改进的基于暗通道先验去雾算法

在雾霾天气下图像质量受大气散射的作用,使得图像质量较差。针对以上问题,提出一种改进的暗通道图像去雾算法。由于带雾图像中局部区域暗通道分量不趋近于0,对大气光值的估计是通过局部像素点R、G、B三通道像素值的方差来判断其波动幅度。若波动较大,则选出三通道中的较小值为当前像素点的暗通道像素值;若波动较小,则通过其周围像素点中最小通道的均值来确定当前点的暗通道值,从而得到精准的暗通道图和大气光值。将采样的方式加快求取透射率图的速度,最后转变为HSI颜色空间可以对图像进行更好的复原。实验结果表明,该算法可以针对色彩失真的状况得到有效的改善,同时在得到去雾图像速度上提高50%以上。

AN IMPROVED PRIOR DEFOGGING ALGORITHM BASED ON DARK CHANNEL

In fog weather, the image quality is affected by atmospheric scattering, which makes the image quality worse. To solve the above problems, we proposed an improved prior image defogging algorithm based on dark channel. Since the dark channel component in the local area of foggy image did not approach 0, the estimation of atmospheric light value was based on the variance of three-channel pixel values of local pixel R, G and B to judge the fluctuation amplitude. If the fluctuation was large, the smaller value of the three channels was selected as the dark channel pixel value of the current pixel. If the fluctuation was small, the dark channel value of the current point was determined by the mean of the smallest channel in the surrounding pixels, and the accurate dark channel map and atmospheric light value were obtained. The sampling method was accelerated to obtain the speed of the transmittance map, and the HSI color space was used to restore the image better. The experimental results show that the algorithm can effectively improve the condition of color distortion and improve the speed of defogging image by more than 50%.