Next: Alternate Missing Columns
Up: Madhav Vyas: Covariance based
In this section I test the given algorithm on textures and on some real and synthetic seismic data sets. I address two different kinds of missing data problems, one where we have alternate missing traces and another where we have large chunks of missing data. Although, the method is not designed for missing chunks of data, as should be clear from the theory, with a few modifications and tricks we adapt the method for that use. The biggest problem for any interpolation scheme is to try to interpolate beyond aliasing; the presented method in its current form fails to accomplish this, but introduction of some intermediate proxy data before filter estimation has improved the results. In this section I discuss all these issues in detail with help of some examples.