How to define a new stochastic process which is function of another process?

I need to define a new process from for example Wiener process like $U(t)=f(W(t))$, (for example $f(x)=1+x^2$ ) and then calculate the average like $\langle U(t)U(s)\rangle$. Is it possible?

-
yes for example simply raising the processes to powers does not work: WienerProcess[]^2 –  richard Jun 2 at 13:31
Comment not an answer, in the documentation pages for the forthcoming Wolfram Language release (and the version of M10 on Raspberry Pi) you can see there is such functionality coming - reference.wolfram.com/language/ref/TransformedProcess.html –  Martin John Hadley Jun 4 at 21:00
@MartinJohnHadley, nice, thanks. –  richard Jun 7 at 9:16

Possibly an overkill for this case but quite general :

define a simple process which returns the variable you want

procU = ItoProcess[\[DifferentialD]x[t] == \[DifferentialD]w[t],  1 + x[t]^2,
{x, 0}, {t, 0}, w \[Distributed] WienerProcess[]] ;


now you can use it as :

Mean[procU[t]]
(* 1 + t *)

CovarianceFunction[procU, s, t]
(* 2 Min[s, t]^2 *)

-