Sparse Unmixing

Sparse unmixing aims to decompose mixed spectral signals, such as those from hyperspectral images, into their constituent pure spectral signatures (endmembers) and their corresponding fractional abundances. Current research emphasizes developing efficient algorithms, including those based on non-convex optimization and archetypal analysis, to address computational challenges and improve accuracy, particularly in semi-supervised scenarios where partial ground truth is available. These advancements are significant for applications like remote sensing, where accurate material identification and quantification from mixed pixels are crucial for environmental monitoring, geological mapping, and precision agriculture.