结合金字塔技术的暗原色先验快速去雾方法  

Fast Removal Method of Haze Combining Dark Channel Priority with Pyramid Technology

作　　者：张丽媛[1] 杨华民[1] 蒋振刚[1] 苗语[1] 张同舟[1] 

机构地区：[1]长春理工大学计算机科学与技术学院,长春130022

出　　处：《长春理工大学学报（自然科学版）》2015年第2期121-124,共4页Journal of Changchun University of Science and Technology(Natural Science Edition)

摘　　要：针对在雾霾天气下无人机航拍的视频图像对比度和色彩保真度差等问题,传统的暗原色先验方法虽然有较明显的去雾效果,但算法复杂度高、耗时长、无法满足实时性需求。本文提出一种基于金字塔技术的快速去雾方法,首先利用高斯金字塔将原始有雾视频图像降采样,然后将降采样后的图像通过暗原色先验的方法去雾处理,最后再采用拉普拉斯金字塔将结果图升采样为无雾视频图像。通过实验结果表明,本文方法在保证去雾效果的同时,提升了场景复原的速度。针对航拍视频场景,能较好地满足实时性要求。In view of video images shot by drones having poor contrast and color fidelity under foggy and hazy weath-ers,traditional dark channel priority method has complex algorithm and is time-consuming,which can not meet the re-al-time requirement,although it has obvious fog removal effects. A fast defogging method based on Pyramid technolo-gy is proposed in this paper.Firstly, the original foggy image will be down-sampled by gaussian pyramid. Then, the sampled images are degogged with the dark channel priority method. Finally, the processed images are up-sampled to fogless video images with the laplacian pyramid. The experimental results show that the method can increase the scene recovery speed on the basis of good defogging effect. For aerial video scene, the method can meet the real-time re-quirement.

关 键 词：航拍视频 暗原色先验 去雾 高斯金字塔 拉普拉斯金字塔 

分 类 号：TN911.73[电子电信—通信与信息系统]