大气光估计-中国图象图形学报.doc

中图法分类号：TP301.6 文献标识码：A 文章编号：1006-8961(200 ) - - 论文引用格式： 基于入射光假设的单幅图像去雾 於敏杰，张浩峰 南京理工大学计算机科学与工程学院江苏，南京 210094 摘要：雾是种常见的天气状况，雾使图像中的景物对比度降低、表面颜色本文提出了一种基于入射光假设的单幅图像去雾方法首先利用全局暗原色进行初步去雾，从而使图像透射率处于[0,1]范围内然后利用雾天光照均匀的特点以及Retinex的照度估计原理进行透射图的估计最后利用透射图以及初步去雾图像得到复原图像。该算法获得的复原图像细节清晰，颜色自然。此外，本文算法也同样适用于灰度图。 关键词：去雾；多尺度；照度估计；Retinex；入射光假设 Single mage dehazing based on dark channel and incident light assumption Yu Minjie, Zhang Haofeng School of Computer Science and Engineering, Nanjing University of Science and Technology, Jiangsu Nanjing, 210094 Abstract: Objective: Fog is a common weather condition, which reduces the foggy image’s contrast, bleaches the surface color, and greatly reduces the value of outdoor images. To solve this problem, we propose a defogging method for single degraded image based on dark channel and incident light assumption. Method: Firstly, in order to find the window which has the maximum mean brightness, we scan the image with a window, and use the average value we got as the atmosphere light. After experiments, we find that the dark channel prior raised by He does not suit for the images that contain large scene. As a result, we weaken the assumption. We assume that there must exist a channel of a pixel, the value of which is zero. Based on this assumption, we are able to find the darkest pixel value in the whole image, and use the darkest pixel value as the global dark channel. After this, we use the ratio of the grayscale of the point to the atmospheric light as the basis transmission of the image. Using the basis transmission, we are able to do the initial dehazing. After doing the initial dehazing, the transmission rate of the image will be stretched to the [0,1] range. Based on the feature we have noticed that images taken under foggy weather almost have no shadow, we assume that incident light in foggy day is uniform. Combined with Retinex theory which uses Gaussian convolution to estimate the illumination, we estimate the transmission by mu