This is a very tricky question to answer. First off, do you want general neuronal density? In that case, choosing a marker that stains neurons (like NeuN) would be helpful. You would have to prepare slices of the hippocampus, preferably very thin (like 20-30µm using a cryostat) and do the actual counting in the regions of interest, counting every Nth slice (N is up to you) as counting all is just impossible.
There's also a way to do cellular counts in slices which does an unbiased estimation of cells. I am personally not familiar with it, but it seems to be called unbiased stereological analysis (something like that was used in an analysis in a paper coming out of my former lab and in their paper the technique used was called optical fractionator method). This is briefly described and referenced in the following article:
Glutamate corelease promotes growth and survival of midbrain dopamine neurons.
Fortin GM, Bourque MJ, Mendez JA, Leo D, Nordenankar K, Birgner C, Arvidsson E, Rymar VV, Bérubé-Carrière N, Claveau AM, Descarries L, Sadikot AF, Wallén-Mackenzie Å, Trudeau LÉ
I believe you would do well to read that and the main reference in the 'methods' section and/or contact Dr Sadikot who I think is the person who has the most knowledge of this technique.
Optical density (Image J) is also usefull for quantify neuronal density in the hippocampus when it is used in association with methods mentioned above.