基于场景深度的雾天图像能见度检测算法

摘要

针对传统能见度测试方法存在硬件成本较高、操作复杂、不能大范围使用等缺陷,提出了一种简单、快速的基于图像的能见度测量方法.首先,对图像进行感兴趣区域的提取、阈值分割和形态学处理.然后,基于大气退化物理模型和暗通道先验原理,估计透射率.再利用同一场景中不同角度所获取的两幅图像提取像素突变点,从而结合摄像机的内部参数估算场景深度.最后,根据能见度检测原理求取大气消光系数,进而获得图像的能见度.试验结果表明,该算法与人眼观测效果一致,准确率较高.与传统方法相比,该方法操作简单、抗干扰性强,不需要任何摄像机标定,并且适用范围广.%The traditional visibility detection methods feature many defects,e.g.,high costs of hardware,complex operation,and small detection range,etc.;to overcome these problems,a simple and fast measurement method of the visibility based on image is proposed.Firstly,the region of interest (ROI) of the image is extracted,and threshold segmentation and morphology treatment are carried out.Then,the transmittance is estimated based on the physical model of the atmospheric degradation and the dark channel prior principle,the pixel mutation point of the two images taken from different angles in the same scene is extracted,thus the scene depth is estimated from the internal parameters of the camera.Finally,the atmospheric extinction coefficient is found according to the principle of visibility detection;and then the visibility of image is obtained.The test results show that the algorithm is consistent with the observation effect of human eye,and the accuracy rate is high.Compared with traditional methods,this method is simple,offering strong anti-interference capability and wide applicable range,even any camera calibration is not needed.