基于暗原色先验的图像去雾算法  

Image dehazing method based on dark channel prior

作　　者：南栋[1] 毕笃彦[1] 许悦雷[1] 王世强[1] 娄小龙[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）》2013年第10期4101-4108,共8页Journal of Central South University:Science and Technology

基　　金：国家自然科学基金资助项目(61175029);国防科技重点实验室基金资助项目(9140c610301080c6106;9140c6001070801);航空科学基金资助项目(20101996009);博士后特别资助项目(2012T50879);博士后基金面上资助项目(2012M512144)

摘　　要：为了减小大气退化现象对可见光成像系统的影响,提出一种优化和改进了的暗原色先验图像去雾算法。该算法从大气退化模型出发,通过自适应分块和中值滤波,得到细化的暗原色图像;利用自适应二维经验模式分解得到照度分量,进而采用低通滤波估计平滑的大气光图像;基于图像的稀疏先验知识,在softmatting下得到细化的大气传输函数图像;并对图像进行视觉色彩校正。最后利用主观和客观的方法对实验结果进行了评价,实验结果表明了算法的有效性。In order to reduce the bad effect of the phenomenon of atmospheric degradation on optical imaging system, an optimized and improved image dehazing method using dark channel prior was proposed. Firstly, median filter and adaptive block was performed with the atmospheric degradation model to get the refined dark channel image. Secondly, the illumination component of brightness image was estimated through adaptive bidimensional empirical mode decomposition, and then smoothened by low-pass filter to estimate the atmospheric optical image. Finally, the refined transmission map was obtained after softmatting based on image sparsity prior, and then the color of image was adjusted by human visual system. The method was evaluated subjectively and objectively with image quality assessment, and the experimental results show the validity of algorithm.

关 键 词：去雾 暗原色先验 二维经验模式分解 色彩校正 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]