基于Curvelet和2DPCA的遥感图像融合算法

A Remote Sensing Image Fusion Algorithm Based on Two-dimensional PCA and Curvelet Transform(1. Collage of Information technology, Pingdingshan University, Pingdingshan 467100, China;

摘要:

根据图像处理不同算法模型的特点,提出了一种基于Curvelet和2DPCA变换相结合的遥感图像融合算法。首先,对多光谱图像进行2DPCA变换,获得其最佳投影轴集合U及特征向量矩阵Q,按照投影规则将多光谱图像投影到U上,得到各主成分分量Yk；再将与多光谱图像进行过直方图匹配的高分辨率图像投影到Q上,获得其主成分PanM及其它主成分分量,将PanM与Yk分别进行Curvelet变换,得到对应的高、低频系数；然后,根据相应的融合规则,对处理后的系数进行Curvelet逆变换,得到融合子图像；最后,将高分辨率图像的其他主成分分量与融合子图像进行2DPCA逆变换得到融合后图像。应用多光谱图像和高分辨率图像进行了融合实验,并将实验结果与其他方法进行比较。实验结果表明,该方法能够在保持源数据光谱特性的同时,较好的提高空间分辨率。

Abstract:

Focusing on the issue that remote sensing fusion image has less spectrum information and low spatial resolution, combining with the characteristics of Curvelet and 2DPCA algorithm, a new image fusion algorithm combining the merits of the two algorithms will be proposed. First of all, perform inverse 2DPCA transform on the multispectral image (MS) to obtain the optimal projection axis U and the eigenvectors Q, then project each band of MS image onto X according to the projection rule to get each principal component.Second, project the Pan image matched histogram with MS image onto Q to acquire the 1st and other principal components, then apply Curvelet transform on the 1st principal components and Yk to obtain the corresponding coefficients.Third, Fuse the coefficients of the decomposed images with proper rules to get the new coefficients. Then perform inverse Curvelet transform on them to acquire the fused sub-image; Finally, perform inverse 2DPCA transform on the fused sub-image and the other components, to get the fused image; The new algorithm has prominent advantage in original image’s spectral information maintenance as well as in improving spatial detail. Experiments carried out on multispectral image and panchromatic high resolution image show that the new algorithm is better than that of former algorithms in comprehensive assessment.