一种自适应的多级透射率估计去雾算法

An Adaptive Multiscale Transmission Estimation Dehazing Algorithm

针对暗通道先验算法在天空区域透射率估计不准确的问题,利用三个不同尺度的高斯函数分别作用于有雾图像的RGB通道来获得"伪"去雾图像;其次,利用有雾图像的混合通道得到自适应参数,将该参数和最小值滤波共同作用于"伪"去雾图像,接着用联合双边滤波消除纹理效应得到透射率的精确估计;最后,采用局部大气光估计方法,结合大气散射模型复原出无雾图像.实验结果表明,该方法不仅降低了时间复杂度,且复原出的图像细节明显,明亮度适宜,对于大面积天空区域有良好的去雾效果,改善了天空区域的颜色失真.

Aiming at the inaccurate transmission estimation in the sky region of the dark channel prior, three Gaussian functions of different scales are used to act on the RGB channels of the hazy image to obtain a "pseudo" dehazing image respectively;Secondly, the adaptive parameter is obtained by using a mixed channel of the hazy image, the parameter together with the minimum filter act on the "pseudo" dehazing image, and then the joint bilateral filter is used to eliminate the texture effect to get the accurate transmission;Finally, using the local atmospheric light estimation method, combined with the atmospheric scattering model to restore the haze-free image. The experimental results show that the method not only reduces the time complexity, but also has obvious image details and good brightness. It also has a good dehazing effect for large-area sky areas and improves the color distortion of the sky area.