Conclusions on my LSJIMP implementation  

In the following section, I discuss issues surrounding my particular implementation of the LSJIMP method.

• Imaging: Imaging of peglegs in this paper is this thesis is accomplished by HEMNO. Prestack migration implicitly scans over unknown, arbitrary reflector dips to remove the effects of wave travel between source and receiver. HEMNO is a single-CMP, analytic moveout equation that assumes known (small) reflector dips. While HEMNO may lack the accuracy and generality of high-end prestack migration methods, it retains convincing advantages in speed and memory usage. Still, improved availability of large cluster supercomputers indicates that least-squares migration methods may soon be feasible, even in 3-D.
• Reflection Coefficient: My modeling of the reflection coefficient of the multiple generator, outlined in Section , is quite simple: a single coefficient. Measuring and applying a higher-order parameterization of reflection coefficient would not be terribly difficult, but the gains might be negligible. My reflection coefficient estimation algorithm benefits from spatial regularization to filter ``noise''; with more parameters, would the smoothing any longer make sense? Also, because LSJIMP is an inversion procedure, the estimated images to some extent will adapt to any unmodeled amplitude variation, though the residual will be biased.
• Other Amplitude Effects: Levin and Shah (1977) perform a detailed analysis of such acquisition-related amplitude effects like source directivity and the response of source arrays and receiver arrays. I have not accounted for any of these effects in this thesis, though such corrections may be straightforward to apply, given sufficient knowledge of acquisition parameters. Nontheless, deficiencies in the modeling are to some extent accounted for by the reflection coefficient.