A Robust Approach to Unsupervised Segmentation of Seismic Data Sets
K. Koester, M. Spann
Proc. VI '98, Vancouver, Canada, June 1998
Abstract
This paper describes an unsupervised method to extract 2D and 3D inner
earth structures from seismic reflection measurements. The application
is a typical texture segmentation problem, which can be split up into a
feature extraction stage and a segmentation stage. As a texture feature,
the locally emergent frequency is estimated by a Gabor filter bank. The
instantaneous frequency (IF) has already been succesfully used for seismic
trace analysis [Taner, 1979] and will be compared with the results of the
filter bank. A second stage of the algorithm involves a region growing
method to compute the final object structure. The extremely flexible segmentation
scheme is appropriate for the application on 2D and 3D data sets of scalar
and vectorial data of arbitrary dimensions. The merging decision is based
on the mutual inlier ratio at the boundary of two adjacent regions. This
ratio is computed by robust regression techniques [Rousseeuw, 1987] to
avoid noise artifacts. A mutual inlier ratio discrimination function to
recognize identical Gaussian distributions guaranteeing a 97.5% certainty
is derived. This method to test distributions is compared with the Kolmogorow-Smirnov-test
and results of the application in a segmentation algorithm are shown. The
segmentation stage is also tested with different data sets from other computer
vision problems to call attention to its general flexibility.
Download gzipped Postscript (887K)
Project
achievements
Back
to papers and internal reports
