In real-world extraction of signal from data we are not given the
needed signal prediction-error filter (PEF). Claerbout has taken
![$S$](img1.png)
,
the PEF of the signal, to be that of the data,
![$S\approx D$](img2.png)
. Spitz
takes it to be
![$S\approx D/N$](img3.png)
. Where noises are highly predictable in
time or space, Spitz gets significantly better results.
Theoretically, a reason is that the essential character of a PEF is
contained
where it is small.