ABSTRACT
Spatial variability often changes with respect to time and the characterization of this change is an important aspect of spatial-temporal modeling. Spatial variation is also important for classification analysis of remotely sensed data since pixels are grouped into classes because of similarities, and classes are distinguished because of dissimilarities. That is, it is necessary to identify and quantify within-band correlations but it is also necessary to quantify and characterize between-band correlation. The use of principal components analysis to remove noise from multiband images is a well-known technique and is based on the premise that the noise term corresponds to a relatively small part of the total variance. Within-pixel and between-pixel variation are related to the total within-image variation. This quantification may be considered for each layer separately or it may be determined for multiband images.
