The Frequency-dividing Multiple Matching and Subtraction Technology Based on Shearlet Transform

It is necessary to have multiple matching during multiple elimination, because there were wide differences between predicted multiples and original records on amplitude and phase caused by the convolution of trace gathers. Conventional multiple matching always use the least-squares matching while the amplitude matching may cause border effect and waveform distortion, and the least-squares subtraction often leads to unsmooth adjacent sample points.Shearlet is a new multi-scale transform with multi-directions, local property, and optimal sparse approximation properties which can be used for presenting multidimensional data.In order to adapt the above-described matching error and obtain more continuous primary event, we combined shearlet transform with multiples matching an use shearlet thereshold subtraction instead of conventional least-squares subtraction. The result shows that this method has better effect.