Projection onto convex sets, or POCS, is a simple, straightforward method of interpolation which hinges on a few basic assumptions. In order to better fulfill these assumptions, a type of POCS is proposed in which a projection onto principal components is the transform used. This method is shown to be effective both on simple synthetic data, as well as on real VSP data. The problem of acquiring the relevant principal components is addressed by using nearby intact data, and other methods are proposed if this should prove impractical. Partially incomplete data are shown to provide adequate principal components for interpolation in some cases.
View full article as PDF (2.05 Mb)