Shadow-Aware Nonlinear Spectral Unmixing With Spatial Regularization

Current shadow-aware hyperspectral unmixing (HySU) methods often suffer from noisy abundance maps and inaccurate abundance estimation of shadowed pixels, as these are characterized by low reflectance values and signal-to-noise ratio. In order to achieve a shadow-insensitive abundance estimation, in this article, we propose a novel spatial–spectral shadow-aware mixing (S3AM) model. The approach models shadows by considering diffuse solar illumination and secondary illumination from neighboring pixels. Besides, spatial regularization using shadow-aware weighted total variation (TV) is employed. Specifically, pixels in the local neighborhood of a target pixel simultaneously consider spectral similarity measures derived from the imagery, elevation similarity measures derived from a digital surface model (DSM), and the impact of shadows. The sky view factor
$F$
, needed as input for the model, is also derived from available DSMs. The proposed approach is extensively validated and compared with state-of-the-art methods on two datasets. Results demonstrate that the S3AM yields superior abundance estimation maps for real scenarios, by decreasing the noise in the results and achieving more accurate reconstructions in the presence of shadows.