Source Separation by Iterative Rank Reduction - Theory and Applications

Blended acquisition is quickly becoming an important topic in exploration geophysics because of the economical benefits it promises. The acquisition technique involves firing two or more sources almost simultaneously, effectively increasing spatial sampling per unit of time, but at the expense of cross-contamination from all the individual source energies overlapping one another on the blended shot record. In this expanded abstract we present an innovative method for separating a blended dataset into its constituent source components. The algorithm is an iterative scheme based on matrix rank reduction, during which the individual source components are estimated. The effectiveness of this deblending algorithm is demonstrated using first a synthetically blended dataset and then a real wide-azimuth blended dataset.