True-azimuth 3D Internal Multiple Attenuation without Subsurface Information

We extend our previous work on 2D internal multiple attenuation without subsurface information to a 3D operation. We describe our implementation that involves selecting traces that honour the azimuth of acquisition for constructing multiple contribution gathers and then segmenting the chosen traces in a layer stripping fashion to predict internal multiple model without identifying the multiple-generating interfaces. We demonstrate through synthetic and field data examples that, by including crossline apertures in the prediction process and selecting traces with correct azimuths in the convolution and correlation processes, substantial improvement in image quality can be obtained for those data that exhibit the internal multiple problem.