顾及图像分割与暗通道先验的去雾霾方法

A Haze Removal Method for Considering the Image Segmentation and Dark Channel Prior

文章针对暗通道先验去雾霾后图像存在颜色失真等问题,提出一种融合图像分割与暗通道先验规律的卫星遥感图像去雾霾方法。首先,引入高斯加权矩阵的梯度算子获取图像的梯度信息,为暗通道去雾霾提供图像分割的约束条件;其次,通过梯度阈值对梯度信息进行划分,从而改善对不同亮度区域的差异化处理;最后,依据梯度阈值修正暗通道透射率来约束暗通道先验的处理结果,实现对不同亮度区域的差异化处理。实验结果表明,无论是主观目视判读还是客观指标对于较为均匀含雾霾影像都具有较好的处理效果。

Aiming at the problem of color distortion in the image after the dark channel prior to removing the haze,a haze removal method for remote sensing image of fusing image segmentation and the dark channel prior rules is proposed in this paper.Firstly,it introduces the gradient operator of the Gaussian weighting matrix to obtain the gradient information of the image,and provides the constraints of image segmentation for the dark channel haze removal.Secondly,the gradient information is divided by gradient threshold,so as to improve the differential processing of different brightness areas.Finally,the dark channel transmittance is modified based on the gradient threshold to constrain the processing results of the dark channel prior and realize the differential processing of different brightness areas.The experimental results show that both subjective visual interpretation and objective indicators have a better processing effect on relatively uniform images containing haze.