Oil and gas exploration decisions are made based on inferences obtained from seismic data interpretation. The interpretation task is getting very time-consuming as seismic data sets become larger. Image processing tools such as auto-trackers assist manual interpretation of horizons-visible boundaries between certain sediment layers in seismic data. Auto-trackers assume data continuities; therefore, their assistance is very limited in areas of discontinuities such as faults.

In this paper, we present a method for automatic horizon matching across faults based on a Bayesian approach. A stochastic matching model which integrates 3d spatial information of seismic data and prior geological knowledge is introduced. The optimal matching solution is found by MAP estimate of this model. A simulated annealing with reversible jump Markov Chain Monte Carlo algorithm is employed to sample from a-posteriori distribution. The model was applied to real 3d seismic data, and has shown to produce geologically acceptable horizons matchings.