基于统计特性和亮度估计的夜晚图像去雾

摘要

由于人工光源的存在,夜晚有雾图像具有光照不均、色偏严重和亮度较低等特点,现有的常规图像去雾算法并不适用.考虑到夜晚雾天图像成像特点,建立了带有色偏因子的雾天图像成像模型,可成功用于夜晚图像去雾,并提出了基于统计特性和亮度估计的夜晚图像去雾算法.由于夜晚雾天图像和低照度图像亮通道直方图具有较大相似性,首先将夜晚有雾图像进行反转,然后基于带色偏因子的雾天图像成像模型,局部估计带色偏的大气光;针对夜晚有雾图像色偏严重,导致三通道透射率差异较大,提出对透射率三通道分别进行处理;在此基础上,利用亮度估计对透射率进行优化;最后,利用局部Grey-world算法对去雾后的图像进行颜色校正.实验结果表明,所提算法能够有效去除夜晚雾气影响,提高图像的整体亮度和对比度,恢复更多的图像细节.%Due to the existence of artificial light,nighttime fog image has the characteristics of non-uniform illumination,color cast and low intensity.The existing image dehazing algorithms are not applicable.Considering the character of nighttime hazy images,a new imaging model for hazy image based on color cast factor was established,which is suitable for defogging of nighttime hazy image.Then the nighttime image dehazing algorithm was proposed based on the statistical properties of nighttime hazy images and the estimation of image intensity.Because of the similarity between the intensity histogram of nighttime hazy images and low light images,the nighttime hazy image was inverted firstly,and then the atmosphere light with color cast factor was estimated locally based on the proposed model.As a resultt of color cast,the three channel transmissions have significant differences,so three channel transmissions were estimated respectively.At last,the transmission was optimized based on the intensity estimation,and the color cast correction was performed using local Grey-world algorithm.Experimental results show that the proposed algorithm can remove nighttime haze effectively,and at the same time the brightness,contrast and image details are improved remarkably.