一种改进的小波变换融合方法及其效果评价  

An Improved Wavelet Transformation Image Fusion Method and Evaluation of Its Fusion Result

作　　者：董张玉[1,2] 赵萍[3,4] 刘殿伟[1] 王宗明[1] 汤旭光[1,2] 刘婧怡[5] 

机构地区：[1]中国科学院东北地理与农业生态研究所,长春130012 [2]中国科学院研究生院,北京100049 [3]安徽师范大学GIS重点实验室,芜湖241003 [4]合肥工业大学资源与环境工程学院,合肥230009 [5]中国地质大学信息工程学院,武汉430074

出　　处：《国土资源遥感》2012年第3期44-49,共6页Remote Sensing for Land & Resources

基　　金：国家重点基础研究发展计划项目课题(编号:2009CB421103);中国科学院战略性先导科技专项子课题(编号:XDA05050101);国家水体污染控制与治理科技重大专项课题(编号:2012ZX07207-004)共同资助

摘　　要：针对传统小波变换融合方法易导致空间纹理信息丢失的缺陷,结合局部方差和局部差异加权算法的优点,提出了一种基于局部算法改进的小波变换融合方法。采用该方法对IKONOS多光谱与全色波段图像进行融合实验,分别从基于视觉效果、数理统计以及面向对象分类精度3个方面分析评价该方法的融合效果。结果表明:改进的融合方法综合了小波变换和局部算法的优点,显著地改善了图像的融合效果,是一种高效的图像融合方法。应用该方法融合后图像的方差由原来的98.28提高到164.32,信息熵由5.30增加到7.85,平均梯度从1.972提高到8.807,图像分类精度提高了10.24%。On the basis of a summary and analysis of wavelet transformation remote sensing image fusion method, in combination with the .advantages of local variance and partial differential weighted criterion, and in the light of the deficiencies of wavelet transform method in enhancing space texture information, this paper has proposed an improved wavelet transformation remote sensing image fusion algorithm. With IKONOS multi -spectral and panchromatic as fusion experiments data, the new algorithm fusion effect was comprehensively evaluated from the subjective, the objective and the object -oriented classification accuracy. The results show that the improved algorithm combined with advantages of the wavelet transform and local algorithm is quite satisfactory. It greatly remedies the defects of traditional wavelet fusion method in remote sensing image texture information loss and serves as a kind of efficient remote sensing image fusion method. With the utilization of the new image fusion method, the remote sensing image variance is raised from the original 98.28 to 164.32, the information entropy increases from 5.30 to 7.85, the average gradient rises from 1. 972 to 8. 807, and the image classification accuracy increases by 10.24%.

关 键 词：图像融合 局部算法 小波变换 面向对象分类 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]