图像雾霾去除方法设计_电子信息工程.rar

摘要:近年来，单幅图像雾霾的去除取得了重大的发展，国内外相关领域的工作者通过成功地使用更好的先验知识和假设，研究出了一些去除图像雾霾的算法。黑色通道先验知识算法(Dark Channel Prior,DCP)首先推算出整体大气光和介质传输率，并通过Softmatting技术优化介质传输率，最后利用大气成像模型得出场景辐射亮度。暗原色先验来自对户外无雾图像数据库的统计规律，它基于经观察得到的这么一个关键事实——绝大多数的户外无雾图像的每个局部区域都存在某些至少一个颜色通道的强度值很低的像素。利用这个先验建立的去雾模型，我们可直接估算雾的浓度并且复原得到高质量的去除雾干扰的图像。对户外各种不同的带雾图像的处理结果表明了DCP算法的巨大作用。同时，作为去雾过程中的副产品，我们还可获得该图像高质量的深度图。

关键词：去除图像雾霾；黑色通道先验知识；大气散射光；介质传输率

Abstract:recent years,the single image haze removal algorithm has made significant development,some researchers has successfully found some image haze removal algorithms by using the better prior and suppose.Firstly,the Dark Channel Prior calculate the global atmospherie light and the media transfer rate.And then optimize the media transfer rate by the Softmatting technology.Finally,get the scene radiance by the atmospheric imaging model.The DCP comes from the statistical regularities of the outdoor fog-free image database,it based on observations such a key facts-the majority of outdoor fog-free image for each local area there are at least one color channel intensity values low pixel.Take advantage of this removing haze by a prior,we can directly estimate the concentration of the fog,and recover high-quality images to remove the fog interference.The outdoor variety of different fog image processing results show the great role of the DCP algorithm.At the same time,as a byproduct of the process to removing haze,we can obtain the high-quality depth map of the image.

Keywords:haze removal,dark channel prior,airlight,transfer rate

通过研究黑色通道先验算法来为单幅图像去雾的实验结果观察来看，黑色通道先验算法是从对户外无雾图像数据库的统计而得出的规律，在雾成像的模型里引入黑色先验，可以使单幅图像的去雾变得更加的有效便捷。通过图11可以看出，黑色先验算法在复原的同时细节部分得以更好的保留，复原图像整体效果较佳。通过限制t(x)的最小边界值，图片中远景处的雾霾被得以部分保留，使得图片在得到复原的同时保留了真实感和距离感。