An Improved Anisotropic Markov Random Field Approach for Prestack Seismic Inversion

This letter presents an improved anisotropic Markov random field (IAMRF) approach for prestack seismic inversion. Instead of using multiple potential functions or adding extra weights on clique potentials as done in existing AMRF approaches, IAMRF removes the effects of anisotropic gradients in subsurface models by means of scaling parameters, which directly tune the model gradients right above. In particular, the scaling parameters are not only directionally varied at certain iterations but also updated by a statistical method at every iteration throughout the inversion, thereby generating more accurate and better edge-preserving inverted results. The prestack seismic inversion based on the IAMRF prior constraints is demonstrated on a field data example, which presents encouraging results.