基于引导滤波器的单幅雾天图像复原算法

Novel algorithm for single haze image restoration based on guided image filtering

基于中值滤波的单幅图像去雾算法所获取的大气面纱图像不能有效地保留雾天图像的边缘信息,也不能真实地反映场景的深度信息,因此,提出了一种基于引导滤波器的大气面纱修正方法。由中值滤波得到初始大气面纱,使用引导图像滤波器对其进行修正得到较为准确的大气面纱,去除多余的纹理信息的同时增强了雾天图像的边缘信息,由大气散射模型得到场景辐射光即复原图像,并对其进行亮度调整。与其他现有的典型去雾算法相比较,该算法在深度剧烈变化的边缘区域有更好的去雾和增强效果,且时间复杂度为线性。

The single image dehazing method based on median filtering neither effectively preserves edges of haze image nor truly reflects the depth of the scene information. So this paper proposes a method of atmospheric veil correction based on guided image filtering. It gets the initial atmospheric veil through median filtering. It is refined by guided image filter to obtain more accurate atmospheric veil. It removes redundant texture information and brings the depth edge information. The scene radiation is obtained by the atmosphere attenuation model, adjusting the brightness of the restored image. Com-pared with other existing typical dehazing methods, the method has a better dehazing effect at distant object where depth changes abruptly, and the time complexity of the methods is linear.