针对图像去雾中由于景深和大气光估计不准确等问题,导致军事监测、目标检测、导航、无人驾驶等系统成像设备获取到的图像质量下降,提出一种结合线性景深估计和自适应雾浓度估计的去雾算法。首先,依照景深与亮度分量和饱和度的关系,利用双滤波优化二者高亮区域,结合线性转换建立线性模型估计景深。然后,提取纹理特征构造雾浓度模型求取自适应散射系数,通过所求景深与自适应散射系数得到透射率。最后,根据对雾图是否含有天空区域的判决,采用两种不同的大气光估计方法。实验结果通过与不同去雾算法定性和定量分析,所提出的方法在保留深度边缘、颜色质量及细节方面具有良好的有效性和鲁棒性,图像恢复质量也相对较佳。

In response to the problems in image dehazing such as inaccurate scene depth and atmospheric light estimation,which lead to the degradation of image quality acquired by imaging devices for military surveillance,target detection,navigation,unmanned vehicles and other systems,a dehazing algorithm combining linear scene depth estimation and adaptive hazy concentration estimation is proposed.First,according to the relationship between scene depth and brightness component and saturation, a linear model is established to estimate the scene depth by using double filtering to optimize the high brightness region of both,combined with linear transformation.Then,the texture features are extracted to construct a hazy concentration model to obtain the adaptive scattering coefficient,the transmittance is obtained through the obtained scene depth and adaptive scattering coefficient .Finally,two different atmospheric light estimation methods are used according to the verdict on whether the hazy image contains sky regions.The experimental results are analyzed qualitatively and quantitatively with different dehazing algorithms,the proposed method has good effectiveness in preserving depth edge,color quality and detail, and the image recovery quality is relatively good.

国家自然科学基金(61561030)、甘肃省高等学校产业支撑计划项目(2021CYZC-04)和兰州交通大学研究生教改项目(JG201928)资助项目