The cost of storing reflection seismic data is enormous and could be reduced by an effective compression technique. While these data may be compressed with several techniques, the need for very accurate reconstruction of the original data is of special importance with seismic data, since details of amplitude and noise are important to later processing.
A simple image compression technique using the singular value decomposition is examined here with reflection seismic data and is found to reproduce the original data poorly because of lost dips and high-frequency detail. This poor reconstruction is caused by the false assumption that the smallest singular values can be ignored with reflection seismic data.
While simpler seismic datasets such as velocity fields and velocity analyses could be successfully compressed using this technique, compressing reflection seismic data is not practical using singular value decomposition, at least not for cases where very-accurate reconstruction is needed.
Examples of compression of a photograph, simple synthetics, a shot gather, and a stacked display are shown.