Coupled Nonnegative Matrix Factorization with Local Neighborhood Weights for Data Fusion

Ertürk A.

2020 Mediterranean and Middle-East Geoscience and Remote Sensing Symposium, M2GARSS 2020, Tunis, Tunisia, 9 - 11 March 2020, pp.41-44 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume:
  • Doi Number: 10.1109/m2garss47143.2020.9105321
  • City: Tunis
  • Country: Tunisia
  • Page Numbers: pp.41-44
  • Keywords: Data fusion, local, NMF, weights


© 2020 IEEE.Unmixing based hyperspectral (HS) - multispectral (MS) data fusion is a relatively recent addition to data fusion literature, and has been shown to provide robust and stable performance. Coupled nonnegative matrix factorization (CNMF) is an unmixing based data fusion method based on alternating unmixing of the HS and MS data while relating the results by point spread function (PSF) and spectral response function (SRF). However, the well-established CNMF method operates solely on the spectral information of the HS and MS data, and disregards the spatial distribution of the data. This paper proposes the integration of spatial information into the update rules used for the abundances in unmixing based fusion under the CNMF framework, based on local neighborhood weights. The proposed approach highlights that the integration of spatial information into the fusion process results in enhanced fusion performance.