At this moment, the main problem that is still not fully solved is that of picking the (Figure 8) distribution. For the examples in this paper, we have artificially constructed this weight from information provided by the forward WEMVA operator, information which we would not normally have when dealing with real data.
In theory, this weight should be consistent with the flatness of the image gathers: smaller values for nearly flat events in the image gathers, and larger ones for events with higher moveout. As we mentioned earlier, we could use other parameters related to this moveout, such as stack power or semblance. In practice, however, any one of these approaches requires picking, which is a very tedious task, especially in 3-D. Perhaps, some kind of automatic picking Clapp (2001) is the answer, but this issue remains to be solved in future work.