联合光谱和空间双尺度细节注入的遥感图像融合  

Remote Sensing Image Fusion Based on Joint Spectral and Spatial Dual-scale Detail Injection

作　　者：王淑香 金飞[1] 林雨准 芮杰[1] 左溪冰 刘潇 杨小兵[2] WANG Shuxiang;JIN Fei;LIN Yuzhun;RUI Jie;ZUO Xibing;LIU Xiao;YANG Xiaobing(Institute of Geospatial Information,Information Engineering University,Zhengzhou 450001,China;232020 Troops of People's Liberation Army of China,Wuhan 430000,China)

机构地区：[1]信息工程大学地理空间信息学院,郑州450001 [2]中国人民解放军32020部队,武汉430000

出　　处：《光子学报》2024年第10期213-224,共12页Acta Photonica Sinica

基　　金：国家自然科学基金(No.42301464)。

摘　　要：设计了一种基于引导滤波增强的联合光谱尺度和空间尺度细节注入的自适应遥感图像融合方法。首先,将全色图像作为引导图像,利用引导滤波对多光谱图像分区域处理,锐化纹理丰富区域、平滑光谱丰富区域,得到增强多光谱图像。其次,利用经典成分替换法和多尺度分析法分别提取光谱尺度和空间差异的细节信息,通过互信息整合得到包含双尺度的细节图像。再次,以原始多光谱图像和亮度分量的光谱相关关系、增强光谱图像的边缘信息作为细节图像的注入约束条件,得到细节注入图像。最后,将多光谱图像与联合空间和光谱双尺度的细节注入图像相加得到融合的高分辨率多光谱图像。分别利用IKONOS、QuickBird、WorldView4和高分二号四类遥感数据进行实验,结果表明,与其他融合方法相比,所提方法在主观视觉效果和客观量化指标上均表现出良好的性能,验证了所提方法的有效性。Remote sensing image fusion is a crucial process that significantly enhances the quality of remote sensing data by integrating multispectral and panchromatic images.However,this integration poses challenges in both spectral and spatial scales.Traditional fusion methods,such as Intensity-HueSaturation(IHS)transformation for Component Substitution(CS)and wavelet transform fusion for MultiResolution Analysis(MRA),each have their own advantages and disadvantages when striving to generate high-resolution multispectral images.The former enhances spatial resolution significantly but leads to noticeable spectral distortion,while the latter maintains spectral information well but has limited spatial resolution enhancement.To fully leverage the strengths of both types of methods,we introduce a novel fusion method that combines spectral scale detail and spatial scale detail.The proposed method consists of three stages.The first stage is the multispectral image enhancement preprocessing,which takes into consideration the different demands for spectral and spatial information of different features in remote sensing images.The panchromatic image is used as a guiding image,and guided filtering is applied to enhance the multispectral image.After enhancement,the texture-rich regions in the multispectral image are sharpened,the gradient's information of which are increased.Meanwhile the spectral-rich regions are smoothed with average filtering to preserve spectral information.The enhanced multispectral image is the foundation for controlling the amount of detail injection in the subsequent stage.The second stage is the detail injection stage,which is a key component of the proposed method and includes detail extraction,detail injection coefficient calculation,and spectral preservation coefficient calculation.Initially,spectral scale difference details and spatial scale difference details are extracted separately from the multispectral and panchromatic images using classical component substitution methods and multi-scale analysis m

关 键 词：引导滤波 互信息 梯度检测 细节注入 遥感图像融合 

分 类 号：P2[天文地球—测绘科学与技术]