Table 1.

Optimal average proportions of endmembers estimated by NLS, NLS with sparsity constraint (S-NLS), NLS with sparsity and low-rank constraints (SL-NLS), and Poisson regression with sparsity and low-rank constraints (SL-PRU).

	AF488	AF514	TET	AF532	AF546	AF555	RRX	AF568	AF594	AF647	AF660	AF680	AF700
NLS	93.1%	60.2%	70.6%	43.6%	79.0%	44.5%	68.8%	61.9%	77.0%	93.9%	64.5%	54.2%	74.4%
S-NLS	97.3%	86.9%	90.0%	62.6%	92.8%	69.7%	82.9%	78.3%	84.9%	98.1%	79.6%	77.3%	81.1%
SL-NLS	98.7%	96.4%	93.2%	72.1%	96.8%	76.8%	91.2%	89.2%	91.3%	99.0%	86.7%	85.4%	87.7%
SL-PRU	$\mathbf{99.5}\%$	$\mathbf{99.5}\%$	$\mathbf{94.5}\%$	$\mathbf{93.6}\%$	$\mathbf{98.6}\%$	$\mathbf{84.1}\%$	$\mathbf{97.8}\%$	$\mathbf{95.0}\%$	$\mathbf{93.7}\%$	$\mathbf{99.2}\%$	$\mathbf{95.7}\%$	$\mathbf{94.4}\%$	$\mathbf{89.4}\%$

Bold represents the best-result for the 4 methods used. The best result was always achieved with SL-PRU.