针对雾天图像对比度差、能见度降低的特点，提出了一种局部尘雾消除方法——LDOS。首先引入暗原色统计先验，说明有雾图像暗通道和其尘雾浓度图之间的近似关系。然后，针对以往的尘雾浓度估计方法存在估计结果粗糙、修正方法复杂、不易实现等问题，给出了一种新颖而有效的估计方法——中值滤波。该方法既能平滑尘雾浓度图，又能很好地保留场景深度发生突变的边缘，使估计出的尘雾浓度图更加精确。最后，在局部区域执行DOS，得到去雾以后的结果。实验结果表明，该方法简单有效，能够较好地达到去雾的目的。

In order to enhance contrast and visibility of scenery images degraded by fog, a Local DarkObject Subtraction (LDOS) method was proposed. According to dark channel prior, the approximate relationship between the fogged dark channel and the fogging density was given. Since the previous methods have the shortages of rough estimated result, complicated correction method and difficult realization, a new and effective median filtering method was presented. The method can smooth the fog density map and preserve large jumps along edges, which may make the estimated fog density map more accurately. Finally, DOS was used in local areas, and the experimental results show that the proposed method is simple and effective for obtaining good results.