Unsupervised segmentation of 3D and 2D seismic reflection data

An unsupervised method to extract 2D and 3D inner earth structures from sei smic reflection measurements is described. The application is a typical tex ture 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 successfully used for seismic trace analysis(21) and will be compared with the results of the filter bank. The second stage of t he algorithm involves a region-growing method to compute the final object s tructure. The extremely flexible segmentation scheme is appropriate for app lication to 2D and 3D images of arbitrary vectorial dimension. The merging decision is based on the mutual inlier ratio of two adjacent regions. This ratio is computed by robust regression techniques(19) to avoid noise artifa cts. A mutual inlier ratio discrimination function to recognize identical G aussian distributions, guaranteeing a 97.5% certainty, is derived. This met hod is compared with the Kolmogorov-Smirnov test and results of the applica tion in a segmentation algorithm are shown. The segmentation stage is also tested with different benchmark data sets from other computer vision proble ms to demonstrate its general flexibility.