Seismic energy that propagates through a complex subsurface can easily be directed outside of the bounds of our survey, causing poor illumination. Additionally, energy that does stay inside the bounds of our survey will be attenuated as some of it becomes evanescent. It is impossible to create a survey of infinite length and we have little control over the formation of evanescent waves, so we have to try to make due with the data we can collect. This data will have areas of poor illumination where noise makes it impossible for us to find the real event. Therefore, we need to try to remove noise and make the event more visible.
Illumination problems generally occur in areas where the subsurface is complex. This means that there are also likely to be artifacts caused by multipathing. To get around this, we choose to work in the angle domain, which doesn't allow multipathing Xu et al. (1998). The methodology we follow to carry out processing in the angle domain has been explained by Prucha et al. (1999).
Reducing multipathing artifacts is not enough to overcome poor illumination. We are trying to improve our model by using a preconditioned inversion scheme. This scheme uses dip penalty filters Clapp et al. (1997) which can be applied in the common midpoint (CMP) - depth plane, the reflection angle - depth plane Prucha et al. (2000), or both.
This paper will explain the theory behind our inversion method and the use of our dip penalty filters. Then we will show the results of using this method on a synthetic dataset and on some real 2-D data. Finally, we will discuss the problems we still must solve and our plans for the future.