中文摘要：

有效地将高空间分辨率全色影像和低空间分辨率的多光谱影像融合起来,对影像上地物的识别具有重要的意义。随着多分辨率分析思想在图像处理中的应用,出现了许多新的多分辨率分解的遥感影像融合方法。典型的方法有：Mallat小波变换（DWT）、àtrous小波变换（SWT）、无下采样轮廓变换（NSCT）,以及这些方法分别与IHS和PCA变换相结合的融合方法。在IKONOS上选取了两个具有不同地物覆盖类型的实验区进行融合实验,从光谱保真度和高频信息融入度两方面,对这些方法进行了定量评价和分析。研究表明,多分辨率分解的影像融合方法均具有较好的高频信息融入度,其中,DWT、SWT、NSCT融合方法在高频信息融入度方面最优;DWT、SWT、NSCT与IHS、PCA相结合的融合方法在光谱保真度方面,不仅优于它们自身,也明显优于传统的IHS、PCA方法,其中,NSCT和PCA变换相结合的融合方法,在光谱保真度方面最强。

英文摘要：

It is of great value to fuse a high-resolution panchromatic image and low-resolution multi-spectral images for object recognition.With the theory of multi-resolution analysis applied in the field of image processing,there emerged many kinds of multi-resolution decomposition-based fusion methods,such as Mallat wavelet transform（DWT）-based,à trous wavelet transform（SWT）-based and nonsubsampled contourlet transform（NSCT）-based fusion methods,or the hybrid of them.The hybrid methods include DWT-I（integrate the DWT with IHS）,SWT-I（integrate the SWT with IHS）,NSCT-I（integrate the NSCT with IHS）,DWT-P（integrate the DWT with PCA）,SWT-P（integrate the SWT with PCA） and NSCT-P（integrate the NSCT with PCA）.These fusion methods were employed in two subsets from an IKONOS image,representing different land cover types.One subset is mainly covered by building,the other is covered by vegetation,water and building.The size of each subset image is 596×596 pixels in panchromatic band.The performance of each fusion method has been further quantitatively analyzed.The mean RMSE（Root Mean Square Error） and correlation coefficients of original and each fused image were used to measure spectral fidelity and high spatial frequency information gain.The study shows that all multi-resolution decomposition-based methods have high spatial frequency information gain,in which the DWT,SWT,NSCT methods are better.PCA-based hybrid methods have better spectral fidelity than IHS-based hybrid methods,but have lower high spatial frequency information gain.All hybrid methods have much better spectral fidelity than the DWT,SWT,NSCT methods and original IHS,PCA methods.The NSCT-based hybrid methods are close to the SWT-based hybrid methods in both spectral fidelity and high spatial frequency information gain.They are much better than DWT-based hybrid methods.The NSCT-P is the best in spectral fidelity.