Recently, Claerbout (1991) and Spitz (1991) have introduced interpolation schemes using a prediction error filter in the time-space domain and in the frequency-space domain, respectively. Both approaches can handle multi-dips at a position, but Spitz's method fails to find prediction error filters when the given data are completely aliased in space. To overcome this limit, Zhang et al.(1992) developed an algorithm which can interpolate in such situations. All these approaches use prediction-error filters to find dips. Although being efficient, they cannot be applied to irregularly sampled traces.
This paper describes an algorithm that combines the advantages of both approaches: the approach of linear interpolation and the approach of prediction-error filter. In the first section, I explain a new algorithm with three steps and the relation to prior algorithms. In the second section, the new method is tested with both synthetic data and real data, which have regular trace sampling interval. The last section describes a way to apply the new method for regridding of irregularly sampled traces and also shows the results of experiments with synthetic data.