Multiple suppression using prediction-error filter

CONCLUSION and discussion

Two synthetic and one real CMP gathers show that this multiple suppression scheme can remove both near-offset and far-offset multiples. This multiple suppression approach has three features. The first one is offset randomization, which destroys the shape of multiple events. The idea of converting coherent noise like multiples into random noise is very novel, it may find applications in other fields. The second is the assumption of primary events being horizontal after normal moveout correction. This assumption makes it difficult in dealing with nonhyperbolic moveout. The third is that this approach cannot handle the amplitude variation along offset. Therefore, it will lose some primary energy caused by AVO or NMO stretch. Obviously, the second and third features will limit the application range of this approach.

I used a trick to improve the efficiency in using T-X-Y PEF. That is, taking the preceding subcube's PEF as an initial guess when estimating a new PEF. In my application, this trick can make the algorithm at least ten times faster than the algorithm without using this trick. The error is less than one percent.

Multiple suppression using prediction-error filter