Abstract :
Abstract - Instruments for microanalysis are now able to
provide several images of the same specimen area. In this
paper, two groups of methods are described for handling
these multivariate maps. One group concerns dimensionality
reduction, i.e., the projection of N-dimensional data sets onto
a M-dimensional parameter space (M < N). It is shown that,
in addition to linear mapping which can be performed by
Multivariate Statistical Analysis, nonlinear mapping can
also be performed (Multi-dimensional Scaling, Sammon’s
mapping, Self-Organizing mapping). The other group
concerns Automatic Correlation Partitioning (ACP). With
these methods, pixels are grouped into several classes
according to the different signals recorded. This can be done
by classical clustering methods (K-means, fuzzy Cmeans) or
by new methods which do not make hypotheses concerning
the shape of clusters in the parameter space.
