Multiple attenuation using eigenvalue decomposition

Multiple reflections constitute one of the most troublesome forms of coherent noise in seismic exploration, especially in marine surveys. There are many approaches to attenuating or suppressing multiples, but none can remove all multiple reflections under all conditions. We have eveloped two new methods to attenuate multiple reflections using the Karhunen-Loeve transform which extract coherent information from multichannel input data in a least-squares sense. They supplement conventional methods. The first is the ‘localized multiple attenuation' method. Its aim is to attenuate a single (typically water bottom) multiple reflection in deep water or multisource marine acquisition. It is similar to velocity filter methods but is not as prone to spatial aliasing and edge effects. The criterion of local coherency gives the method the flexibility to discriminate between primary and multiple reflections. The second is the ‘targeted multiple attenuation' method. It uses the similarity between a primary reflection and its multiple in both the spatial and time directions. Its aim is to remove a particular multiple reflection associated with a specified primary rather than all the multiple energy for an entire trace. The method can be applied pre- or post-stack, and to 2D or 3D data. Keywords: multiple, reflection, suppression, eigenvector, eigenvalueGlobal Journal of Pure and Applied Sciences Vol. 12(2) 2006: 225-22