Abstract: Images acquired in bad weather have poor contrasts and colors. This paper proposed a simple method to remove haze based on dark channel priority. After acquiring the transmission, getting the difference between the dark channel and dark value of nearest eight pixels, the pixel of minimal difference was redefined as new dark channel. Besides, the air light was automatically estimated from the histogram of the dark channel. At last, the clear image could be recovered based on physical model. The experimental results show that the method can sharp the edge and improve the quality of the degraded image.

Key words: dehazing, physical model, dark channel prior, transmission, neighborhood similarity

摘要： 针对雾天图像对比度和颜色严重退化的问题,提出了一种应用暗原色先验去雾的简便方法。该方法首先根据暗原色先验得到雾天透射率图;然后,通过暗原色值与邻域暗色值的差值来判断边缘,取颜色最相近点的暗色值为新的暗原色值,同时利用暗原色图像直方图自动获得大气光线值;最后,根据大气散射模型得到清晰图像。实验证明,该方法能够有效去除雾对图像的影响。

关键词: 去雾, 物理模型, 暗原色先验, 透射率, 邻域相似性