改进HOT的高分影像自动去薄云算法  

Automatic Removal of Thin Cloud Based on Improved Haze-Optimized Transformation for High-Resolution Image

作　　者：刘泽树[1,2] 陈甫[2] 刘建波[2] 孙业超[3] 

机构地区：[1]中国科学院大学,北京100049 [2]中国科学院遥感与数字地球研究所,北京100094 [3]中国资源卫星应用中心,北京100094

出　　处：《地理与地理信息科学》2015年第1期41-44,F0003,共5页Geography and Geo-Information Science

基　　金：民用航天十二五预研项目

摘　　要：为了有效去除薄云对高分光学影像造成的干扰,解决传统算法自动化程度低、去薄云效果不理想的问题,提出改进的薄云最优化变换(Improved Haze-Optimized Transformation,IHOT)算法。首先采用暗原色先验知识从薄云影像中自动选取晴空区,运用薄云最优化变换检测薄云,再创新性地利用植被区域云检测精度较高的特点改进检测结果,最终使用虚拟云点法进行薄云去除。利用高分一号影像进行实验,证明该文提出的算法能够有效地去除薄云对高分影像的影响,尤其对人造地物的色彩和纹理信息恢复效果优于传统算法。A new algorithm is developed to remove the haze from high-resolution image effectively and improve the automatic de- gree of cloud detection methods. The clear region of haze image is selected automatically based on the prior of dark channel, and the thin cloud detection is generated by improved haze-optimized transformation （IHOT）, then the haze is removed with the vir- tual cloud point method, Using the GF-1 image to make experiment, the proposed algorithm is proved to be better than the tradi- tional algorithms in recoverin~ the color and texture information of artificial objects.

关 键 词：高分影像 去薄云 自动检测 IHOT 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]