城市航空影像中基于颜色恒常性的阴影消除  

Removing shadows from urban aerial images based on color constancy

作　　者：叶勤[1,2] 徐秋红[1,3] 谢惠洪[4] 

机构地区：[1]同济大学测量与国土信息工程系,上海200092 [2]同济大学遥感与空间信息技术研究中心,上海200092 [3]同济大学现代工程测量国家测绘局重点实验室,上海200092 [4]上海市测绘院,上海200063

出　　处：《光电子．激光》2010年第11期1706-1712,共7页Journal of Optoelectronics·Laser

基　　金：国家"863"计划资助项目(2007AA12Z178)

摘　　要：基于颜色恒常性理论,对真彩色和彩红外城市航空影像中高大建筑物形成的阴影进行消除。首先采用光谱比技术和最大类间方差法(Otsu)阈值分割技术进行城市航空影像中建筑物阴影的检测,进而就颜色恒常计算的Shades of Gray算法中明可夫斯基范式(Minkowski norm)的p取不同值情况下的阴影去除效果进行实验,利用亮度、对比度及平均梯度值比较阴影去除效果的好坏。实验表明:在基于航空影像阴影区域及非阴影区域划分的基础上,本文方法比一般的阴影区反差拉伸方法效果好;且与一般场景影像的阴影去除不同,对两类航空影像,p取2时阴影去除效果最佳,说明这两类影像不能简单看成是一个灰色世界影像。A method is used to remove high building shadows in true color and color infrared urban aerial images based on the theory of color constancy.The spectrum ratio and Otsu threshold segmentation methods are used to detect building shadows in true color and color infrared urban aerial images.Then, based on the shadow detection result,one of the color constancy algorithms shades of gray（SoG） is used to remove the shadows in aerial images with different p values of Minkowski norm.Finally, the shadow removal results with different p values are compared with brightness, contrast and average grads.The experiments show that the result of this method is better than that of the contrast stretching method in the shadow area based on the segmentation of shadowed and unshadowed areas.And different from the result of the general scene image shadow removal.The aerial images get the best results of shadow removal when p is 2.It means that the two types of aerial images should not be simply regarded as a gray world image.

关 键 词：航空影像 颜色恒常性 阴影消除 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置]