基于邻域窗口滤波的图像去雾

Image Dehazing Based on Neighborhood Window Filtering

户外摄影成像质量常常受雾霾等能见度低的天气影响,为了提高雾天的成像质量,提出了一种基于邻域窗口滤波——NWF(Neighborhood Window Filtering)的图像去雾算法。在大气散射模型下使用暗原色统计先验,大气透射率和暗通道有关,直接使用暗通道估计出的大气透射率结果粗糙,还原的图像边缘有光晕看起来不够真实。针对这一问题设计了一种简单高效的邻域窗口滤波器从八个邻域方向上计算切尾均值自适应选择出最佳的方向来提取精细的大气透射率图,该方法不仅能平滑尘雾浓度图,又能从根本上保留纹理信息。结合大气散射模型使用具有更多纹理边缘信息的的透射率图,最后可以还原出一幅更加真实的去雾图片。实验结果表明,该方法相比于传统的导向滤波及双边滤波的去雾方法更加简单,效果明显。

The quality of outdoor photography imaging is often affected by weather with low visibility such as haze.In order to improve the imaging quality of foggy days,this paper proposes an image defogging algorithm based on Neighborhood Window Filtering(NWF).This method is based on the atmospheric scattering model and the dark channel prior technique,and uses a neighborhood window filter with smooth texture preservation function to restore high-quality pictures.The feature of edge-preserving filtering using the domain window filter can extract more transmittance maps with texture edge information from the dark channel prior model.Combined with the physical model of atmospheric scattering,it can finally restore a high-quality defogging picture.The experimental results show that the method is simpler and more effective than traditional guided filtering and bilateral filtering defogging methods,and the effect is obvious.