Hypermix is a new open-source tool for remotely sensed hyperspectral image unmixing. It includes several popular algorithms covering different steps of the hyperspectral unmixing chain:

  • Estimation of the number of endmembers:
    • Hyperspectral subspace identification algorithm (Hysime)
    • Virtual dimensionality (VD)
  • Endmember extraction.
    • Orthogonal subspace projection (OSP)
    • N-FINDR algorithm
    • Vertex component analysis (VCA)
    • Spatial-spectral endmember extraction (SSEE)
    • Automatic morphological endmember extraction (AMEE)
    • Spatial pre-processing (SPP)
  • Abundance estimation.
    • Linear spectral unmixing (LSU)
    • Non-negative constrained linear spectral unmixing (LSU)
    • Sum-to-one constrained linear spectral unmixing (SLSU)
    • Fully constrained linear spectral unmixing (FCLSU)

In addition, the tool includes techniques for dimensionality reduction and for quantitative and comparative evaluation of the results of spectral unmixing using reference spectral signatures in a library and other metrics such as the root mean square error (RMSE) in the reconstruction of the original scene using the results provided by the linear unmixing process. For additional details, please read the following paper.