基于细节增强的单一图像去雾

Single image de-haze based on detail enhancement

在雾霾、烟雾等恶劣天气状况下拍摄出的照片能见度很差,对视野中物体的可分辨性很低。在暗通道先验的基础上描述了一种有效增强细节的单幅图像去雾算法。首先对雾霾图像进行边缘保持模糊处理,进而提取出不含雾成分的细节层。与此同时对雾霾图像使用中值滤波算法去雾,将初步去雾后的图像加上细节层,得到更加细致、清晰的图像。为了提取图像的细节层,选用了域变换递归边缘保留滤波器,它能够在保持边缘的同时快速对图像进行模糊处理。

Images of outdoor scenes show poor visibility and discernibility in the inclement weather conditions such as haze, smog and mist. This paper describes a new method for single image de-hazing which can enhance the details effectively. Based on the dark channel prior, it first blurs the fogged images with edge-preserving filtering, and then extracts the detail layer without fog. In the de-hazing process of the foggy image, the detail layer is appended to the preliminary de-hazing image produced by the median filter to get more meticulous and sharper images. In order to extract the detail layer of foggy image, it chooses the domain transform recursive edge-preserving filter, which can perform high-quality edge-preserving filtering of foggy images in real time.