Reducing Change of Illumination Effect for Hyperspectral Image Unmixing

In the field of the processing of hyperspectral images, the pixel mixture is a serious problem to resolve. This difficulty comes from several outliers which affect seriously the reliability of spectral unmixing results. The illumination change effect, where the image do not reflect the true appearance of the scene in many cases due to primarily by slope or shadow facts, is considered one of the most important outliers and it is essential to deal with this problem which can otherwise have a serious effect on the estimation results. Several attempts have been made to correct this problem which is not currently handled by atmospheric correction: these approaches allow the Endmembers of the mixture to be random variables (mostly Gaussians) and lack the ability to explain the statistical variability of the spectra within a class. The present work propose a new approach called Independent Component Analysis and Spectral Angle Measure based Spectral Unmixing (ICA-SAMSU) which use the spectral angle constraint for abundance quantification. The major benefit of this approach is its capability to estimate abundance quantification independently of the amplitude (magnitude) of the spectral signatures, using only spectral angle measures. As a consequence, a significant reduction in spectral unmixed error corresponding to the spectral similarity within-class confusion is obtained. A second benefit concerns physical constraints which are respected. The experiment was conducted using simulated and real image in order to validate our approach and to compare it with a well known statistic one.