一种新的单幅图像快速去雾算法  

A new fast single image dehazing algorithm

作　　者：娄小龙[1] 毕笃彦[1] 李权合[1] 南栋[1] 

机构地区：[1]空军工程大学航空航天工程学院,陕西西安710038

出　　处：《中南大学学报（自然科学版）》2014年第6期1854-1859,共6页Journal of Central South University:Science and Technology

基　　金：国家自然科学基金资助项目(61175029)

摘　　要：针对图像雾霾退化的恢复,及现有算法实时性差的缺陷,提出一种高效的单幅图像去雾算法。该算法从雾霾天气条件下图像退化模型出发,首先通过对图像进行白平衡,简化图像退化模型,并消除大气光对恢复图像色彩的影响,提高色彩恢复的准确度,而后通过暗通道先验求解大气传递图的粗略估计。根据大气光衰减项的先验知识与假设,通过构造罚函数对传递图进行约束,采用规范映射最陡下降法直接求解退化模型中的大气光耗散项。最后将大气光耗散项带入退化模型反解出场景反照率。实验结果表明:该算法能够较好的实现对退化场景图像色彩恢复与细节信息的增强,较好地提高退化图像的视见度,且算法复杂度低,处理速度快。A new fast image dehazing algorithm was proposed for the purpose of the single image dehazing based on the model of degradation. Firstly, the white balance was performed in order to simplify the scattering model and the accuracy of the result was improved. And then the dark channel prior was utilized to estimate the rough transmission of the image. The prior knowledge and the prior assumptions of the airlight were used to construct the penalty function. And the airlight of the degradation model was obtained utilizing the PNSD. Finally, the scene albedo was recovered by the model. The experimental result shows that the algorithm can recover the color and restore the detail of the image effectively and improve the visibility of the image with a fast implementation. Besides, the algorithm is simple.

关 键 词：图像去雾 暗通道先验 罚函数 二次规划 视见度 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]