为了提高多光谱图像的空间分辨能力的同时更大程度地保持光谱信息, 提出了结合多元经验模态分解和加权最小二乘滤波器的遥感图像融合方法.多元经验模态分解解决了传统的基于单变量经验模态分解的遥感图像融合方法中多光谱图像的亮度分量和全色图像分解出的子图像频率不匹配导致融合图像空间细节信息缺失问题, 加权最小二乘滤波器可以精确地估计出源图像的低频信息继而得到高频信息, 减小了传统的经验模态分解方法估计的高频信息中混有低频成分而导致的光谱失真问题.将两者的优点结合, 采用不同的融合规则得到的融合图像在空间细节和光谱信息的保持度较好.选取多组不同卫星数据进行仿真实验, 并与结合多元经验模态分解和àtrous小波变换的方法以及基于加权最小二乘滤波器的遥感图像融合方法等方法进行比较, 实验结果表明本文方法在光谱质量和空间分辨率方面都取得了很好的性能.

In order to improve the spatial resolution of multispectral images while maintaining spectral information to a greater extent, this paper proposes a remote sensing image fusion based on multivariate empirical mode decomposition and weighted least squares filter. On one hand, multivariate empirical mode decomposition solves the problem of spatial information distortion caused by the subimage frequency mismatch between the intensity component of the multispectral image and the panchromatic image in traditional remote sensing image fusion methods based on univariate empirical mode decomposition. On the other hand, remote sensing image fusion based on multivariate empirical mode decomposition usually suffer from serious spectral distortions due to the detail information contains low frequency components. To overcome these defects, the weighted least squares filter can estimate low frequency information of source image accurately and obtain the high-frequency information subsequently. Combine the advantages of both, the fused image obtained by different fusion rules has better spatial detail and spectral information retention. In this paper, different satellite data are selected for simulation experiments, and compared with other methods such as based multivariate empirical mode decomposition and àtrous wavelet transform and based on weighted least squares filter, the results of experiment achieve good performance in both spectral and spatial qualities.