The field data is a 3-D marine seismic sail line from the North Sea. The survey system consists of two sources and three streamers. There are many conflicting dipping reflections and high frequency diffraction hyperbolas in the dataset, which provide a critical challenge to this compression technique. In this application, we compress the dataset in two ways: before NMO correction (case A) and after NMO correction (case B). The size of this dataset is n1=500, n2=112, n3=6, and n4=150. We compress/decompress the two different kinds of dataset at three compression ratio levels, 40, 70, and 100. Figure 14 is the original dataset. Figures 15 and 16 are the results of CompRatio=40. Figures 17 and 18 are the results of CompRatio=70. Figures 19 and 20 are the results of CompRatio=100.
As shown in the figures, the result of after NMO correction is better than that of before NMO correction for each of the three compression levels. This is because NMO correction can strengthen the extent of coherency in the dataset. The SNR of the whole dataset is higher in case B. We choose the result of CompRatio=70 to analyze the change of SNR in the time domain, frequency domain, and wavenumber domain, as shown in Figures 21, 22, and 23. In the time domain, the shallow zone result of post-NMO correction is prominently better than case A. In the deep zone, which is mainly composed of high-frequency uncorrelated noise, the SNR has decreased greatly in both cases. In the frequency and wavenumber domain, they show similar results, i.e., the result of case B is better than that of A in the low-frequency and low-wavenumber areas. In the high-frequency and high-wavenumber ranges, the accuracy of this compression algorithm is lowered.