非负矩阵分解算法在遥感图像融合中的应用

NON-NEGATIVE MATRIX FACTORIZATIONS APPLIED TO IMAGE FUSION IN REMOTE SENSING

研究了非负矩阵分解(NMF)方法在遥感图像融合中的几种应用.在多光谱图像与全色图像融合的过程中,采用了非负矩阵分解融合算法,非负矩阵分解与主成分结合(N_PCA)的融合算法,非负矩阵分解与提升小波变换结合的融合算法,通过对各融合图像的目视判定及统计参数判定,分析评价这些算法在遥感图像融合中的性能差异.研究实验证明非负矩阵分解算法应用于遥感图像融合处理,具有较好的融合效果,非负矩阵分解算法,非负矩阵分解与主成分结合的融合算法,非负矩阵分解与提升小波变换结合的融合算法在遥感图像融合中的性能优于传统的主成分融合算法(PCA),其中,非负矩阵分解与提升小波变换结合的融合算法的性能最为优异.

Multi-spectral images have good multi-spectral features in spite of its low spacial resolution, with little details of the earth＇s surface information. In contrast, panchromatic images have high spacial resolution but limited spectral features. The features of the two can be fused into fusion images, which will make it easier to evaluate or process by computation. In the present paper, several new non-negative matrix factorization （NMF）-based fusion algorithms are presented for such fusions of multi-spectral and panchromatic images, including NMF and principal component transform （PCA）-based fusion （N-PCA）, NMF and lifting wavelet transform （LWT）-based fusion （N-LWT）. The potentials of each are summarized. It was found that NMF applied to image fusion in remote sensing improved the quality of fusion images. A comparison of all proposed new fusion methods with PCA shows that the former is better than the later, N-LWT fusion is the best of the lot.