基于暗通道先验和Retinex理论的快速单幅图像去雾方法

针对雾霾天气下捕获的图像存在低对比度、低饱和度和色调偏移等现象, 提出了一种基于暗通道先验和Retinex理论的快速单幅图像去雾方法.该方法从大气散射模型出发, 利用暗通道先验法则,通过灰度开运算对大气光值进行区间估计,同时获得介质传输率的初始估计, 并通过白平衡简化大气散射模型; 其次,基于Retinex理论,利用高斯滤波获得介质传输率的粗略估计, 并通过线性映射实现灰度值搬移; 然后,将介质传输率的初始估计和粗略估计进行像素级融合, 利用快速联合双边滤波进行边缘优化,同时通过参数自适应调整的方法对雾图中大片天空区域的介质传输 率进行修正; 最后,通过简化大气散射模型和色调调整得到复原图像.与几种典型的图像去雾算法相比, 本文算法具有很快的运算速度,能有效提高复原图像的清晰度和对比度,同时获得较好的图像颜色.

A Fast Single Image Dehazing Method Based on Dark Channel Prior and Retinex Theory

Imaging in the atmosphere presents the phenomenons of low contrast, low saturation and hue offset due to atmospheric particles such as haze and fog. In this paper, a fast method is proposed to remove haze from a single image based on dark channel prior and Retinex theory. Based on the atmospheric scattering model, dark channel prior and gray-scale opening operation are used to estimate the value of global atmospheric light by an interval. Meanwhile, the initial estimation of medium transmission is obtained. And the white balance is performed to simplify the atmospheric scattering model. Then, a simple Gaussian filter is adopted to get the coarse estimation of medium transmission based on Retinex theory, whose gray value is altered by linear mapping. After the pixels fusion between the initial estimation and coarse estimation of medium transmission, a fast joint bilateral filtering is used to refine edge, and a dynamic parameter strategy is adopted to improve the medium transmission of large sky region in fog image. Finally, the simplified atmospheric scattering model and tone mapping are used to get the restored image. Compared to some state-of-the-art methods, the proposed method can achieve a faster processing speed, effectively improve the visibility and contrast of the restored image, and obtain good color effect.