基于特征的遥感图像信息融合模式研究  

Research on model of remote sensing image information fusion based on feature

作　　者：付炜[1] 邢广忠[1] 侯蓝田[1] 王文君[1] 

机构地区：[1]燕山大学信息学院通信与电子工程系,河北秦皇岛066004

出　　处：《测绘科学》2004年第6期62-64,共3页Science of Surveying and Mapping

基　　金：国家自然科学基金资助项目(69662001)

摘　　要：基于图像特征的遥感图像信息融合是在突出目标地物的空间结构和纹理特征情况下的信息融合。本文在数字图像小波多分辨率分析理论基础上,采用小波变换方法对高分辨率遥感图像的目标地物边缘进行信息增强,然后与多光谱遥感图像进行特征信息融合。在融合过程中,首先对多光谱图像中的R、G、B三个波段的图像进行小波分解,得到相应的低频图像,并对特征增强后的高分辨率图像进行小波分解,再将分解后的高频图像分别与低频图像进行融合,最后经RGB合成为彩色图像。该方法既改善了图像的清晰度和分辨率,同时也保留了原图像的光谱信息。本文最后通过融合实验验证了上述结论。The image feature fusion is a kind of data fusion that is based upon spectral structure and texture feature of the objects. This paper proposes feature fusion approach after enhancing Remote Sensing (RS) image edge with wavelet's multiresolution analysis,and largely improve image's definition and resolving power. In the fusion process, instead of using the conventional HISWavelet fusion, the author used low frequent parts of the R,G,B bands of the multispectral image which is decomposed by the wavelet and high frequent parts of the highresolution image which is also decomposed by the wavelet to do the image fusion separately, and finally the new fusion image is composed by the new R,G,B bands. In this way, the methodology not only improves the definition and resolution of the multispectral image, but also retains the color feature which is important for multispectral fusion. At last the author validates the conclusion mentioned above with an experiment.

关 键 词：小波变换 特征融合 边缘增强 

分 类 号：TP752.4[自动化与计算机技术—检测技术与自动化装置]