Statistics of hyperspectral data/image analysis: Entropy

In this work, we are focusing on the development of metrological features for the analysis of hyperspectral measurements. We propose to define the entropy of a hyperspectral dataset or image. As the processing cannot be developed in the spectral acquisition space, the proposed entropy is processed using histograms of spectral differences. The metrological properties are induced by an adapted spectral difference, the Kullback-Leibler pseudo-divergence. Results are proposed for different color scales and show how the entropy can be used to select spectral references for hyperspectral image analysis or processing.