An Overview on Linear Unmixing of Hyperspectral Data

Table 1

The pros and cons of some sparse unmixing methods.

Author	Year	Pros	Cons
Sigurdsson et al.	2016	This algorithm can be applied on very large hyperspectral image where a standard algorithm could not be used due to memory limitations. And each subproblem does not need to have access to the whole hyperspectral image.	Endmembers in the scene are assumed to be the same with its corresponding sample in the spectral library; this assumption is too ideal in the real environment.
Wang et al.	2017	To solve the existing spectral variability between the measured endmembers in the real environment and corresponding ones in the spectral library.	The method only uses spectral information but ignores the spatial-contextual information.
Sigurdsson et al.	2017	A novel sparse outlier component is used to remove outliers and structured noise from the unmixing algorithm.	The algorithm is easy to make the estimated data appear over smooth.
Ruyi et al.	2017	This paper adopts Bregman divergence for sparse unmixing, which is a differentiable, smoother prior. And based on the maximum a posterior estimation, the proposed method has achieved sparse, stable, and precise fractional abundances.	Endmembers in the scene are assumed to be the same with its corresponding sample in the spectral library, and this assumption is too ideal in the real environment.
Zhenwei et al.	2018	l0 norm was used to solve l 0 problem directly. The antinoise ability of this algorithm also gets improved.	This algorithm is difficult to integrate into a larger learning architecture.
Huang et al.	2019	This paper developed a two-level reweighting strategy to enhance the sparsity along the rows within each block and considered spatial correlation among nearby pixels.	If this algorithm will be applied on very large hyperspectral image, the computation time will be too long.