97.

[Cover] [Contents] [Index]

Page 185

measures other than the root mean square error, and Robinson et al. (2000) give details of results of spectral unmixing applied to images with differing spatial and spectral resolutions.

One of the problems in spectral unmixing is that it is often difficult to define the spectra of an adequate set of end members. This can be a restriction in cases in which the geographical distribution of a limited number of end members is of interest. Nielsen (1998) describes a method of partial spectral unmixing, in which the effects of a small number of end members can be estimated without the need to specify the complete end member ensemble. See also Harsanyi and Chang (1994), Miller et al. (1992) and Settle (1996).

[Cover] [Contents] [Index]