Hyperspectral image fusion by multiplication of spectral constraint and NMF

Hyperspectral remote sensing has been used in many fields, such as agriculture, military detection and mineral exploration. Hyperspectral image (HSI), despite its high spectral resolution, has lower spatial resolution than panchromatic image (PI). Therefore, it is useful yet still challenging to effectively fuse HSI and PI to obtain images with both high spectral resolution and high spatial resolution. To solve the problem, a new HSI fusion method based on multiplication of spectral constraint and non-negative matrix factorization is proposed in the paper. In the model, the HSI is first decomposed into basis (abundance matrix) and weight (spectral matrix), then the details of HSI are sharpened by enhancing the details of the abundance with PI. Meanwhile, a spectral constraint term is proposed. It is used to specifically preserve the spectral information in the model. Therefore, the fused data is characterized by good spatial and spectral information. Finally, experiments with both simulated and real data are implemented and the results show that the proposed method performs better in both visual analysis and objective indices than conventional methods, thus making it a good choice for HSI fusion.