基于改进暗通道先验的图像去雾算法  

Image Dehazing Algorithm Based on Improved Dark Channel Prior

作　　者：张宏[1] 张玉伦 邓旭 徐梅 ZHANG Hong;ZHANG Yu-lun;DENG Xu;XU Mei(College of Computer Science&Technology,Harbin University of Science and Technology,Harbin Heilongjiang 150080,China)

机构地区：[1]哈尔滨理工大学计算机科学与技术学院,黑龙江哈尔滨150080

出　　处：《计算机仿真》2022年第4期150-155,共6页Computer Simulation

基　　金：黑龙江省大学生创新创业训练计划项目(201710214011)。

摘　　要：针对雾霾天气下采集的图像存在细节丢失和对比度较低的问题,提出了基于改进暗通道先验的图像去雾算法。现有的图像去雾算法仍然受到颜色失真和亮度偏暗的困扰。而改进算法首先通过四叉树搜索算法来估计大气光候选区域,提高了大气光值估计的准确性,降低了大气光候选区域定位在亮白色物体上导致去雾后的图像出现颜色失真的情况。其次,将去雾后图像转为HSI颜色空间,对亮度进行限制对比度自适应直方图均衡化处理,有效的改善了去雾后图像亮度偏暗的问题,并且更加凸显图像的细节。最后,在真实有雾图像上的实验结果表明,改进算法具有较好的去雾效果并且亮度更自然。To solve the problems of detail loss and low contrast in images collected under haze weather, an image dehazing algorithm based on improved dark channel prior is proposed. The existing image dehazing algorithms still suffer from color distortion and dark brightness. The improved algorithm estimates the candidate regions of atmospheric light by the quadtree search algorithm, which improves the accuracy of atmospheric light value estimation and reduces the atmospheric light candidate regions fall on bright white objects, resulting in color distortion in the dehazed results. In addition, the dehazed image was transformed into HSI color space, and the brightness I was processed with contrast limited adaptive histogram equalization. This effectively improves the dark brightness problem of the dehazed image and highlights the details. Finally, the experimental results on real hazy images show that the proposed algorithm has a better dehazing effect and more natural brightness.

关 键 词：图像去雾 四叉树搜索 暗通道先验 颜色空间 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]