Abstract: When a UAV is in the aerial photography mission, it is easily to be affected by some factors such as fuselage shaking, terrain environment and etc., which will lead to blurred images, and will greatly influence the results of subsequent image information extraction. Inorder to solve this problem, the automatic detection method of aerial image quality based on re-blur theory is proposed. Firstly, the original image which is preprocessed by compression, graying and etc. attaches a certain degree of Gaussian blur to get re-blur image. Then, the edge images are extracted by Laplace operator respectively from the two images, and the edge difference image of these two edge images are obtained. Standard deviation of the edge difference image is calculated to compare with the threshold value which is used to divide sharp and blur image, deriving from experience, determining whether the image is a blurred image. Finally, the experiment is carried out on the synthetic image and the actual image. Experimental results show that the algorithm has a high detection rate of blur image, which is superior to other image quality detection algorithms, and the speed of image detection is very fast.

Key words: image quality detection, human visual perception, re-blurred theory, UAV aerial image, edge difference, blur detection

摘要： 无人机在进行航拍任务时，会因为机身抖动、地物环境等原因导致采集的图像模糊，对后续提取图像信息造成影响。针对这一问题，提出了一种基于再模糊理论的无参考图像质量检测方法，用来区分清晰和模糊图像。对原始图像进行缩放、灰度化等预处理后附加一定程度的高斯模糊，得到再模糊图像，再分别对两张图像使用拉普拉斯算子提取边缘，得到两张图像的边缘差异图像。通过计算所得的边缘差异图像的标准差与经验得出的划分清晰和模糊图像的阈值相比，判断该图像是否为模糊图像。对人工合成的模糊图像和无人机实拍图像进行实验。实验结果表明，该算法具有较高的模糊图像检测率，表现优于其他图像质量检测方法，且单张图片的检测计算速度很快。

关键词: 图像质量检测, 人类视觉感知, 再模糊理论, 无人机航拍图像, 边缘差异, 模糊检测