Forghani,
I have seen very similar behavior in capacitors as a result of dielectric absorption (see Modeling dielectric absorption in capacitors at www.designers-guide.com/Modeling). Ferromagnitism is the dual of dielectric absorbtion and so has a similar effect. Effectively what is happening is that at low frequencies the magnetic domains have more time to line up with the magnetic field imposed by the winding, which supplements the field and hence the inductance. The time constants of the domains are widely distributed and so you see this gradual increase in the inductance as frequencies decrease.

Turns out, the same is true for skin effect (see Modeling skin effect in inductors also at www.designers-guide.com/Modeling). With skin effect, the impedance exhibited by an inductor is roughly
Z = jwL + sqrt(jw)/H
and if you define
L_eff = Z/(jw)
then
L_eff = L + 1/(sqrt(jw)H)
which increases slowly as w --> 0.

Which effect is to blame depends on the frequency range and type of inductor. I suspect in your case it is ferromagnitism that is the culprit.

When measuring the inductance, be sure to also measure the real part of the inductor's impedance. It is key to being able to extract and build an accurate model.

-Ken