非负矩阵分解在遥感图像融合中的应用

Application of Non-negative Matrix Factorization to romote sensing image fusion

非负矩阵分解(Non-negative Matrix Factorization,NMF)算法是在矩阵中所有元素均为非负数的条件下的一种矩阵分解方法,这为矩阵分解提供了一种新的思路。非负矩阵分解方法在图像处理领域具有十分重要的应用意义。介绍了非负矩阵分解的基本思想,讨论了非负矩阵分解用于图像融合的可能性,并实现了基于非负矩阵分解的遥感SAR图像与SPOT图像的融合,NMF能通过观测图像数据找到图像的基矩阵,发现图像的特征,从而最终获得融合图像。不仅对基于NMF的融合方法进行了实验,而且对基于NMF的融合方法和基于小波的融合方法作了对比,并从主观和客观上来评价了这两种融合图像的质量。实验结果表明基于NMF的融合图像与原始的SAR图和基于小波的融合图像相比,能提供更多的信息,更适合作为实时定位的基准图。

Non-negative Matrix Factorization （NMF） is a kind of matrix decomposition method with the constraint that each element of matrix is nonnegative,which has wide application in image processing.In this paper,the basic principle of NMF is first introduced,then the application of NMF in image fusion is discussed,and at last,the fusion of SAR image and SPOT image based on NMF is implemented.NMF can find basis vector through observation data,and discover image feature,obtaining the fusion image.This paper not only studies on the image fusion method based on NMF,but also makes comparison between it and fusion method based on wavelet transforms,evaluating these fusion images quality with evaluation criteria.The experiment results show that fusion image based on NMF can offer more information than original SAR image and fusion image based on wavelet transforms,more suiting to use as reference map for real-time location.