Not really sure why you would ever need to try to "annualize" covariance with respect to rates of returns.
Any reason you'd have to do this?
Statistically COV(cX,kY) = ck*COV(X,Y)... in your case, if X and Y are monthly and you "annualize" them, you'd have to multiply your covariance by 144.
Similarly VAR(cX) = c^2*VAR(X)