改进的暗通道先验图像去雾霾算法

Improved dark channel prior image haze removal algorithm

对图像去雾问题进行研究,提出一种改进的暗通道先验图像去雾霾算法。先根据图像亮度最高处的像素值计算大气光值,然后在获取透射率的同时参考峰值信噪比,利用迭代的方法求出使峰值信噪比最优时所对应的透射率,并采用导向滤波的方法细化透射率,最后通过有雾图像物理模型还原图像。实验结果表明,相比于经典的暗通道去雾算法,改进算法能够显著改善图像信噪比以及大面积天空图像出现的光晕问题。

To study the problem of image haze removal,an improved dark channel prior algorithm for image haze removal is proposed.The atmospheric light value based on the pixel value at the highest brightness of the image is calculated first.Then,the transmissivity is obtained with reference to the peak signal-to-noise ratio(PSNR).The transmissivity at the optimal PSNR is obtained by iterative method,and the transmissivity is refined by guided filter algorithm.Finally,the image is restored by the physical model of foggy image.The experimental results show that compared with the classic dark channel algorithm,the proposed algorithm can significantly improve the peak signal-to-noise ratio and solve the halo problem of large area sky image.