Here we introduce a new method for 5D seismic data interpolation that is based on low rank radial function interpolation. Being different from traditional methods that exploit the sparsity of data in Fourier/Curvelet/TauP domain, and use an iterative subtraction approach to fit the data, our method uses the linear combination of a set of radial functions to fully represent the seismic data. The advantages of this method are: first, our method guarantees full representation of the original data thus no "information loss" which iterative approach cannot achieve; second, our method is not based on Fourier transfer thus result does not have artifacts caused by Gibbs phenomenon; Finally, the implementation of this method involves only matrix operation thus can be easily parallelized and migrated to GPU or FPGA devices. We use a model data example and a field data example to demonstrate the quality and effectiveness of our method.
展开▼