摘要
针对遥感图像能量低的实际问题,提出了一种亮度直方图局部线性化图像增强方法来提高彩色遥感图像的可视效果。首先,对RGB模型描述的彩色遥感图像进行HSI变换,以有效分离H、S和I分量;其次,对亮度I分量进行传统的直方图均衡化,得到均衡化灰度映射曲线;然后,将图像梯度作为目标函数,求出最优的线性化折点位置,对灰度低端动态范围映射曲线进行线性化处理,得到局部线性化的灰度映射曲线;最后利用新的灰度映射曲线对图像进行增强处理。Himawari-8真彩色图像增强实验结果表明,经亮度直方图局部线性化增强后,像素平均梯度由73提高到了147,较传统的RGB域直方图均衡化的123及HSI域直方图均衡化的134高;图像信息熵由5.87提高到6.63,全部优于传统的RGB域直方图均衡化和HSI域直方图均衡化。本文方法有效地改善了彩色遥感图像的可视效果,提高了图像对不同目标的辨识能力。
Aimed at actual problem of low energy of remote sensing image,a local linear image enhancement method of luminance histogram was put forward to improve visual effect of color remote sensing image.Firstly,HSI transformation was performed on color remote sensing image described by RGB model to separate component H,S and I effectively;secondly,traditional histogram equalization was implemented to luminance I component and gained equalized gray level mapping curve;then,by taking image gradient as objective function,the location of optimum linear break point were determined and linear processing was conducted to mapping curve within dynamic range at low side of gray level and gained local linear gray level mapping curve;finally,new gray level mapping curve was adopted to enhance image.Experimental result of Himawari-8true color image enhancement shows that after local linear enhancement of luminance histogram,level gradient of pixel increases from 73to147,and with traditional RGB domain histogram equalization,it only increases to 123,with HSI do-main histogram equalization,it increases to 134,and information entropy of image increases from 5.87 to 6.63,so it can be conclusion that the proposed algorithm is superior than other two algorithms in all aspects.The method improves visual effect of color remote sensing image effectively and improves identification capability of image to different objectives.
作者
陈博洋
CHEN Bo-yang(National Satellite Meteorological Center, Beijing 100081, China)
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2017年第2期502-508,共7页
Optics and Precision Engineering
基金
国家自然科学基金面上项目(No.41375023)
关键词
遥感
图像增强
HSI模型
直方图均衡化
remote sensing
image enhancement
HSI model
histogram equalization