Semi-realistic simulations of natural hyperspectral scenes

Many papers in the hyperspectral literature use simulations (based on a linear mixture model) to test algorithms which either estimate the dimensionality of the data or endmem-bers. Typically these simulations use (i) 'real world' end-members, (ii) proportions distributed according to a uniform or Dirichlet distribution on the endmember simplex, and (iii) Gaussian errors which are 'spectrally' and 'spatially' uncor-related. When the error standard deviations (SDs) in different bands are assumed to be unequal, they are usually estimated using Roger's method. The simulated and real world data in these papers are so different that one can't be confident that the various advocated methods work well with real world data. We propose a methodology which produces more realistic simulations, providing us with greater insights into the strengths and weaknesses of various advocated methods. In particular, using an AVIRIS Cuprite scene, we demonstrate that Roger's SD estimates are positively biased.