As illustrated in Figure 4, shot gathers of the pseudo-primaries are similar to those of the original dataset. We now migrate the pseudo-primary dataset by shot-profile, downward-continuation migration. We use Fourier finite difference Ristow and Ruhl (1994) as our wavefield extrapolation operator for the migration. Figure 6 compares the image from the primary-only dataset (NFS), the image from the original dataset (FS), and the image from the pseudo-primaries, which consists of primaries (from first-order multiples) and multiples (from higher-order multiples). Figure 6a shows the migration result of primaries, Figure 6b shows the migration result of the original dataset and Figure 6c shows the migration result of pseudo-primaries. The pseudo-primary image is noisier than the images from the primaries only and the original dataset. This is caused by the noise from the cross-correlation and first-order multiples in the pseudo-primary gathers. In the image obtained by migrating the pseudo-primaries (Figure 6c), the salt body is clearly imaged, and reflectors below the salt are also well imaged. However the reflectors below 20,000 ft are contaminated by first-order multiples in the pseudo-primaries, which are similar to those in the original-dataset image (Figure 6b).