基于局部学习的遥感图像融合

本文提出了一种基于局部学习的遥感图像融合方法。兵基本思想是在局部区域划融合图像与全色图像建立对应的局部线性关系。由于图像数据在局部区域相对简单，因此局部模型相比全局模型更为合理。在局部学习的基础上，将全色图像与融合图像的全局回归误差表示为图拉普拉斯的形式，其本质是利用局部学习使得融合图像保持全色图像的流形结构。同时为了保持多光谱图像的性质，通过图像的尺度空间表示，建立融合图像与多光谱例像之间的尺度关系。最后通过集成融合图像的二次拉普拉斯形式和足度空间表示，构建图像融合的全局目标函数。为了优化目标函数，本文提出了闭合求解法和快速迭代求解法。实验结果表明：本文所提出的融合方法比传统融合方法具有更好的效果。

Remote Sensing Image Fusion Based on Local Learning

This paper presents a local teaming based method for remote sensing image fusion. The key idea of our method is to construct the local linear model between the panchromatic image and the fused image in local region. Compared with the complexity in global region, image data in local region are much simpler. This means that local model is more reasonable than global model. Based on local learning, the regression error between the panchromatic image and the fused image can be formulated with Laplacian quadratic form, which makes the fused image preserve the manifold structure of the panchromatic image. In order to preserve the properties of the multispectral image, we build the scale space relationship between the fused image and the multispectral image via scale space representation. Consequently, the objective function of our model is proposed by combining the Laplacian quadratic form and the scale space representation. Meanwhile, a close tbrm solution method and a fast iterative method are proposed to optimize the proposed model. Experimental results demonstrate the effectiveness of our model in comparison to the state-of-the-art methods.