Reverse-time migration (RTM) provides superior images in areas where there are steep salt flanks or other complex geologic structures. However, the high cost of running RTM with regard to memory requirement and computation time makes it difficult to use RTM for routine large volume production. By dividing the subsurface into two or three regions in depth according to the structures of the velocity model and applying RTM from top to bottom sequentially in each region, we are able to make RTM very cost effective for production usage. Furthermore, Kirchhoff migration or one-way wave equation migration may be used to replace RTM in a region where the velocity model is relatively simple and RTM may not help to generate a better image. This hybrid approach may further improve the computation efficiency and the quality of migration images.
Wavefield extrapolation (WFE) multiple prediction typically operates in common shot domain, and is very effective to predict complex multiples such as diffracted multiples; however, it is a model-based and relatively expensive approach. In this paper, we present a new WFE multiple prediction algorithm which operates in common-P domain. This new algorithm is able to very efficiently predict source-side multiples, and less model-dependent; therefore enables its usage in the early stage of processing. In this paper, we first perform some analysis of two types of diffracted multiples : 1) source-side diffracted multiples and 2) receiver-side diffracted multiple. With the new insight of source-side diffracted multiples, we have developed a new algorithm to predict source-side multiples. This new algorithm for source-side multiple prediction is a variant of WFE approach, but almost model independent. Testing on both synthetic and field data demonstrates the effectiveness of the new multiple prediction method.