CONCLUSIONS

Using different datasets, we test the performance of Chevron's seismic data compression software. The results show that it is a very promising approach. Especially this new technique provides a possible solution to the contradiction between the increase of 3-D seismic dataset and the trend toward distributed and networked computing enviroments. The most prominent characteristic is this method's high compression ratio, which is achieved in the high-dimensional wavelet domain.

Although this compression method is a general technique, there are still some empirical rules that will help us to attain better results:

• The conducting of some tests on small-sized dataset to find an acceptable compression ratio and the corresponding scale value. Since there is no fixed quantitative relation between scale value and compression ratio, same scale value can result in different compression ratios when applied to different datasets. Even at the same compression ratio, the compression quality (for example, SNR) also varies from one dataset to another. Therefore, it is necessary to perform some tests and compare the corresponding results to select an appropriate scale value.
• The organization of the dataset into different groups to facilitate the compression. In order to achieve a better result, we should make use of the coherency of the dataset as much as possible. The reasonable organization of the dataset is essential to the exploration of the coherency. Usually, for 3-D datasets, the common-shot gather and common-midpoint gather are both acceptable input forms for the compression session.
• The application of some conventional processing, such as NMO correction and AGC gain, to the original dataset prior to carrying out wavelet transform. As shown in last section, NMO correction can enhance the coherency of the original dataset and improve the performance of this compression technique. Thereafter, applying some processing procedures which can strengthen the coherency is undoubtedly beneficial to the final result. But there is a trade-off between the time spent by preprocessing and the effect achieved by the same procedure.