线性约束下的高斯自适应标准差去雾算法

Gaussian Adaptive Standard Deviation Dehazing Algorithm Under Linear Constraint

针对暗通道先验算法中透射率在天空等明亮区域估计过小以及最小滤波使用不足问题,提出一种线性约束下的高斯自适应标准差去雾算法.首先利用有雾图像的最小通道图构造高斯函数来近似估计无雾图像的最小通道效果,提升明亮区域透射率的准确度;为了防止无雾图像最小通道灰度级超出范围,提出线性系数进行约束,使得其灰度级分布在(0,1);再通过观察得到高斯函数标准差与雾浓度呈负相关,提出自适应标准差来控制最终复原效果,并利用交叉双边滤波进行优化;最后结合大气散射模型复原图像.在Matlab R2014a软件中进行仿真实验,并采用视觉效果分析和盲评估评价指标对文中算法进行验证.结果表明,与经典算法对比,该算法有效地恢复了有雾场景的对比度和细节信息,明显增加了图像可见度且具有较低的时间复杂度.

Aiming at the problem that the transmission of dark channel prior algorithm is too small in bright areas such as the sky area and the minimum filter is insufficient,a Gaussian adaptive standard deviation dehazing algorithm under linear constraint is proposed.Constructing a Gaussian function by using the minimum channel map of the hazy image to approximate the minimum channel effect of the haze-free image,thereby improving the accuracy of transmission in bright areas such as the sky area.In order to prevent the gray level of minimum channel of the haze-free image from exceeding the range,a linear coefficient is proposed to constraint so that the gray level is distributed within(0,1).Secondly,it is observed that the standard deviation of Gaussian function is negatively correlated with haze concentration,so that an adaptive standard deviation is proposed to control the final restoration effect,and optimized by cross-bilateral filtering.Finally,the hazy-free image is restored by combining the atmospheric scattering model.The simulation experiments were carried out in Matlab R2014a software,and this algorithm were verified by visual effects analysis and blind evaluation indicators.Numerous experimental results show that compared with the classical algorithm,this algorithm effectively recovers the contrast and detail information of the hazy scene,which obviously increase the image visibility and has lower time complexity.