We apply spatial contiguity analysis (SCA) to study spatial structures cont
ained in seismic images. Compared to classical methods, such as principal c
omponent analysis (PCA), SCA is more efficient for multivariate description
and spatial filtering of this kind of images.
We present SCA according to geostatistic formalism defined by Matheron. A p
reliminary spatial analysis of initial variables is required. Made with the
help of variogram curves, this permits to underline spatial properties of
these variables and defines contiguity distance and direction to apply SCA.
A series of mathematical tools is defined. They allow quantifying the infor
mation held by initial variables and factorial components in terms of varia
nce and spatial variability and exhibit data spatial structures on differen
t scales.
The method is applied to analyse a seismic data set, We compare PCA and SCA
results. This data set gives us the opportunity to show the interest of pr
eliminary spatial analysis of initial variables, and the effects of spatial
direction and distance on the data decomposition in elementary structures.
(C) 2001 Elsevier Science BN. All rights reserved.