空间和光谱信息保持的多光谱图像融合算法

Pan-sharpening algorithm based on spatial and spectral information preservation

采用基于空间和光谱信息保持的多光谱图像融合框架,以生成具有高空间质量的多光谱图像.融合框架的能量泛函包括4项:边缘自适应提取约束项、线性组合系数约束项、光谱信息保持约束项和波段比例关系保持约束项.前两项旨在提高融合图像的空间质量,后两项旨在减轻图像的光谱失真程度.将这4项能量泛函综合成一个总能量泛函,并利用欧拉-拉格朗日公式和梯度下降法对其求解,同时调整正则化项,以使得融合图像的空间质量和光谱质量都能得到明显的改善.在QuickBird和Pavia University图像数据上进行仿真实验,结果表明,与SFIM、MTF_GLP、MTF_GLP_HPM、PCA、GS、GSA、AIHS、GFPCA等算法相比,本文方法的融合图像具有较高的空间和光谱质量.

A pan-sharpening algorithm based on spatial and spectral information preservation was proposed to generate multispectral images with high spatial quality.The proposed method consisted of four energy functionals:the edge adaptive extraction constraint term,linear combination coefficient constraint term,spectral information preservation constraint term and band proportion relationship preservation constraint term.The first two terms were aimed at improving the spatial quality of the fused image,and the latter two terms were aimed at reducing the spectral distortion of the fused image.The four energy functionals were integrated into a total energy functional,which was solved by Euler-Lagrange formula and gradient descent method,and the regularization terms were adjusted to improve the spatial and spectral quality of the fused image.Simulation experiments carried out on QuickBird and Pavia University datasets demonstrate that compared with SFIM,MTF_GLP,MTF_GLP_HPM,PCA,GS,GSA,AIHS and GFPCA,the proposed method can produce fused images with higher spatial and spectral quality.