You already have some nice answers here and there, but I'll also post mine here.
UPDATE: For more information you can have a look here.
Mark is right in the fact that the brain can adapt (for more information see this and this publications). But in my opinion this is not directly connected to the question here. I do believe that you should try to "correct" the frequency (and time) characteristics of your headphones for "optimal" reproduction.
Without going into much details here, one way to "cancel" the effect your headphones "impose" to the signal, is to calculate the inverse filter for your headphones. This is not always easy and even more, not always the optimal solution. It can introduce time smearing, excess ringing and even non-flat frequency response if not done correctly.
One simple way is to try to invert the frequency response of your headphones like Tetsujin proposed in his comment. Although, as already mentioned do that with increased caution, as it may result in unwanted artifacts making the result worse than the original version.
I don't really support the idea of using graphical equalizers for such processes (like Polyterative suggest..., I believe), as most of the time their centre frequencies and Q factors (bandwidths) don't give enough flexibility to work with (most of the time you will end up treating frequencies that you don't have to). They may end up providing an improvement though, if you are lucky enough (so that centre frequencies and bandwidths come close to the "problematic" regions of your frequency response).
Ideally you would measure the response of your headphones with some miniature microphones at the entrance of your ear canal and then try to invert that. This has of course some complications starting with "how am I supposed to measure that?" and going as far as to "what is the inverse of the response?".
If you have a measurement laboratory close to you, or an audiologist has appropriate equipment, and you are willing to go into the process, you could measure your headphones there. There are also implications here as you should somehow also take care of the blocked ear canal induced alterations but this is kinda minor compared to the "imperfections" of the transducers I believe.
Regarding the inversion, it is a process that needs some technical expertise, which involves filter design and calculating inverse filters. You could achieve that with tools like MATLAB and its Signal Processing Toolbox or its free and open-source counterpart Octave and Signal Package. Otherwise, if you want something less technical, you could possibly use any frequency analyzer that would give you both magnitude and phase.
Finally, possibly the best (and safest?) solution would be to get in contact with the manufacturer and ask for the inverse filter, or just the frequency response (or its inverse) and use that.