The document discusses adapting the IR-MAD change detection method for use with hyperspectral data. It proposes using principal component analysis (PCA) for feature reduction before applying IR-MAD to address the high dimensionality of hyperspectral data. An initial change mask is also used to eliminate strong changes and isolate no-change pixels for analysis. Experiments on Landsat and hyperspectral data demonstrate the effectiveness of the approaches. Current work involves using Markov random fields to incorporate spatial information and generate a final change classification map.