Seismic Signal Deconvolution Using a Maximum Posterior Mode Algorithm and a Gibbs Sampler

In this paper we introduce a multichannel blind deconvolution algorithm for the restoration of two-dimensional (2D) seismic data. This algorithm is based on a 2D reflectivity prior model, which takes into account the spatial dependency between neighboring traces. Each reflectivity column is estimated from the corresponding observed trace using the estimate of the preceding reflectivity column and a modified maximum posterior mode (MPM) algorithm. The MPM algorithm employs a Gibbs sampler to simulate realizations of seismic reflectivities. We apply the algorithm to synthetic and real data, and demonstrate improved results compared to those obtained by a single-channel deconvolution method.