Hi gladiator,
you gave more explanation but stil it is not clear enough what is the frequency range you have in mind. If you have to use such high dielectric constant, I assume the operating frequency is relatively low. In this case, I am not quite sure that coupled line filters are the best choice for you (not saying that they are impossible, just might not be the optimal solution). You might be better off using lumped components instead.
High dielectric constant substrate will certanly decrease the size of the filter as electrical length is inversly proportional to square root of effective dielectric constant, and it is again proportional to relative dielectric constant of substrate.
You can think on Er effect like it is making the field lines more dense. For the same reason in order to couple resonators to achieve desired filter response, space between resonators must be small. Depending on substrate thicknes, precise fabrication of small gaps might be a problem, thus filter will be different from (any) simulations and will have wide spread of characteristics in production.
I suggest as thick substrate as you can get to make line and gaps reasonably wide. Length of resonators should not change much compared to thin substrate, so you will have your small dimensions and resonable tolerance requirements.
Good luck
flyhigh