Abstract: An adaptive single image enhancement algorithm is proposed to deal with serious image degradation in haze, dust, rain, snow and other extreme weather. First, an image classifier is designed to judge whether the image is degraded. If so, different methods are taken to process the image separately according to chrominance components. Second, the transmittance of haze images is obtained by segmenting the bright area based on dark channel prior algorithm, which improves halo phenomenon occurred in restored image. What’s more, this method is extended to snow images. For dust images, Gamma correction is introduced in limited contrast adaptive histogram equalization algorithm to correct contrast and brightness imbalance problems. Finally, compared with other algorithms, this method improves the clarity of the image and also avoids the generation of halo. It solves the contrast and brightness imbalance problem in dust image enhancement.

Key words: haze removal, rain and snow removal, dust removal, Gamma correction, image classifier

摘要： 针对雾霾、雨雪、沙尘等极端天气下获得的图像严重退化的问题，提出一种自适应的单幅图像增强算法。首先设计一种图像分类器，判断图像是否为降质图像，若是则根据色度分量值对图像分别处理。其次对于雾霾图像，在暗原色先验算法基础上，通过分割图像的明亮区域求取透射率，改善了原算法复原的图像易产生光晕的现象，并将该算法扩展应用于雨雪图像；为了处理沙尘图像，采用限制对比度自适应直方图均衡化算法，为了校正该算法处理图像时对比度和亮度失衡的问题，采用伽马校正。与其他算法对比实验表明该算法有效提高了图像的清晰度，同时避免了光晕的产生，解决了沙尘图像处理中对比度和亮度失衡的问题。

关键词: 去雾霾, 去雨雪, 去沙尘, 伽马校正, 图像分类器