一种导向滤波优化的暗原色先验去雾算法

A Priori Dehazing Algorithm for Dark Primary Colors Based on Guided Filtering Optimization

大气中的残存颗粒和雾霾会对摄像设备造成干扰,导致图片模糊,影响图片清晰度,对图像进行复原具有较强的现实意义。基于暗通道先验去雾算法的原文中利用软抠图(soft matting)进行信息图的优化,但在试验中发现,当像素点的强度值接近大气光值时,去雾的图像会出现局部的色斑、色偏效应;另一方面原算法太耗时,平均对一张图片去雾所耗费的时间都在40s以上;基于以上问题,本文提出了一种引入导向滤波(guided filter)对信息图进行优化的暗原色先验去雾算法算法。实验结果表明,算法的整体执行速度得到大幅提升,对于色彩反差较小的图片,其去雾效果有了明显的改善,同时,该算法还有良好的边缘平滑特性。

Residual particles and haze in the atmosphere will cause interference to the camera equipment,thus resulting in blurred images,so restoring images has a strong practical significance. In the original text based on the dark channel prior defogging algorithm,soft matting is used to optimize the information map. However,experiments show that when the intensity value of the pixel is close to the atmospheric light value,the defogging image will have local color spots,color shift effect;on the other hand,the original algorithm is time-consuming,and the average time it takes to defog a picture is more than 40s. To solve these problems,this paper proposes a dark primary color first that introduces guided filtering to optimize the information graph verify the dehazing algorithm. Experimental results show that the overall execution speed of the algorithm is greatly increased. For images with small color contrast,the defogging effect is significantly improved. At the same time,the algorithm has good edge smoothing characteristics.