Gradient Based Spectral Similarity Measure for Hyperspectral Image Analysis

Spectral matching algorithms or similarity measures dis-criminate spectral signatures of unknown materials by comparing them with the reference spectral signatures. Spectral matching algorithms play an important role in identifying different earth surface features using hyperspectral image analysis. Identification of diagnostic features and discriminating spectral signatures, especially of mineral signatures, which exhibit subtle differences among themselves is still a challenging task. Thus, developing or coming up with a new spectral matching algorithms is expected. Therefore, in this paper, we present a gradient based spectral similarity measure (GSSM) that captures the diagnostic (absorption) features to measure the degree of closeness between spectral signatures. Effectiveness of the proposed GSSM in distinguishing spectrally distinct signatures is studied with that of other spectral matching algorithms in (i) discriminating endmember signatures (ii) mixed pixel identification (iii) clustering spectral signatures of different classes and (iv) endmember extraction. Experimental results on a spectral library of 5 minerals formed using a benchmark mineral dataset called Cuprite data clearly show that the proposed GSSM is capable of (i) discriminating different spectral signatures in a better manner and also (ii) bringing into notice of the user the locations of diagnostic features by highlighting them to recognize easily.