Texture analysis and segmentation of seismic images

A method is proposed for the texture analysis and segmentation of geophysical images. It is based on the detection of the seismic horizons and on the calculation of their features (e.g. length, average reflection strength, signature). These features represent the texture of the seismic image. The horizons are clustered into classes according to one or several of their features. Each cluster represents a distinct texture characteristic of the seismic image. After this initial clustering, the points of each horizon are used as seeds for geophysical image segmentation. All pixels in the seismic image are clustered in those classes, according to their geometric proximity to points lying on classified horizons. Thus the entire seismic image is classified on the basis of seismic texture patterns. Two methods are proposed for clustering pixels according to their geometric proximity to reference points. The first is based on Voronoi tesselation and mathematical morphology. The second is based on a so-called radiation model for region growing. Simulation examples are presented