At the higher frequencies you probably depend on
filter caps for most of the PSRR. The error amp loop
has the usual gain rolloff and can't help you much.
Input filter C can help if the input supply has a
reliable inductive characteristic. But now you are
getting application specific. Input filter ESR/ESL
does set a maximum filter effectiveness. N caps
of 1/N value in parallel will be a lot stiffer (low
ESL and ESR) than one big one. This may not
cost you much, or anything.
A RC, LC, RLC input filter could work if, after its
I*R drop, the supply at the LDO remains within
tolerable range. This wouldn't impact load regulation
the way an output complex filter would. But transient
performance might be compromised if the C is too
small to hold up.
Many LDOs have trouble with very low ESR output
filter caps, they need a built-in zero or get unstable.
That goes against what you want. PMOS LDOs have
a lot of trouble with positive supply noise due to the
PMOS final device's gain. You might find one of the
ULDOs using NMOS pass devices have better HF
PSRR, although these need a second auxiliary supply
higher than the power path source - got one?
I have seen a few LDOs from Linear Tech which are
pretty proud of their HF PSRR and rate it out to,
if I recall, 40MHz? Whether the rating satisfies,
would be your call.