INTRODUCTION

Most conventional data processing treats missing data the same as it treats zero-valued data. Improving this is a problem of long standing. Potentially the most rewarding circumstances for such research is when data contains strong events that can be extrapolated significantly beyond the bounds of measurement. The extended events need not be signal--they can be coherent noise such as multiple reflections. By extrapolating such events we sharpen the velocity spectrum, thus better outlining the velocity of all reflections.

A variety of optimization strategies have been proposed in the literature with none particularly established in practice. Here I test the idea of minimizing the relative power measured in velocity space. Unfortunately, the final result casts doubt on the formulation that follows below.