High Quality Horizon Mapping Using Clustering Algorithms

In this paper, we present a clustering based method used to process 3D seismic data and automatically map seismic horizons in the presence of discontinuities. Our approach uses the cosine of instantaneous phase attributes and applies Principal Component Analysis to the original datasets of trace shapes to improve the quality of the original samples. We also propose a measurement to infer the quality of the clusters used to map the seismic horizons. Based on this measurement, we show that using the cosine of instantaneous phase attributes and PCA greatly improves the mapping of seismic horizons.