利用目标和大气偏振信息的雾天图像重构方法

结合目标偏振信息和大气偏振信息的差异, 提出了一种利用目标和大气偏振信息的雾天图像重构方法。首先, 从光强图像中分离大气光图像和目标光图像, 分别解析大气光偏振信息和目标光偏振信息, 构建偏振去雾模型。然后, 采用融合图像梯度信息的高斯滤波方法估算大气光强和目标光强, 并分别计算大气光偏振度和目标光偏振度。采用3σ法则阈值分割方法, 在大气光图像空间内估算无穷远处大气光强。最后, 重构出目标图像。在不同天气环境下开展外场实验, 结果表明, 该方法在雾霾、雨、雪天气下能够较好地恢复出目标信息, 重构后图像的信息熵提升约40%, 灰度标准差提升了约90%, 平均梯度和边缘强度提高了3倍。

Fog image reconstruction using target and atmospheric polarization information

Combining the difference between target polarization information and atmospheric polarization information, an image reconstruction method based on the polarization information of target and atmosphere was proposed. Int this method, firstly, the atmospheric light image and target light image are separated from the light intensity image, the atmospheric light polarization information and target light polarization information are analyzed respectively, and the polarization defogging model is constructed. Secondly, the Gaussian filtering method fused with image gradient information is used to estimate the atmospheric light intensity and target light intensity, the atmospheric light polarization degree and target light polarization degree are calculated respectively.Then, the threshold segmentation method of 3σ rule is used to estimate the atmospheric light intensity at the infinite distance in the atmospheric light image space and finally the target image can be reconstructed. Field experiments were carried out under different weather conditions.The experimental results show that the method proposed can recover the target information well in fog, rain and snow weathers, and the entropy of reconstructed image increases by about 40%, the standard deviation of gray level increases by about 90% and the average gradient and edge strength increase by 3 times.