Ok, so you want to control the filter instead of making a DSP processor using the NIOS II core. In this case I would next determine the hardware structure of the filter. Getting the numbers from the NIOS to the hardware will not be difficult so I think the actual DSP portion of the project will be the challenge (not sure that will make things easier or harder for you).
The lookup table aspect has me a bit confused since I would just program registers that feed the multipliers (just send the coefficients to the NIOS processor and have it program your filter using the "Interface to User Logic (IUL)" in SOPC builder) but perhaps you have some size constraints that need to be taken into consideration. I also noticed that you are using a Cyclone FPGA, are you aware that it does not have dedicated DSP hardware? So every multiplier or MAC you have in your design is going to require quite a few resources. So to answer your question of where to start I think the filter itself would be the first step. In Matlab filters cranked out with 50 coefficients are nice, but if you try to implement a true FIR chain of these in an FPGA you will run out of room fast. Instead I would determine the sampling rates and the computational times you need and try to share logic resources.
<div class='quotetop'>QUOTE </div>
--- Quote Start ---
The filter should work on 100Mhz if possible. The AD converter is connected to the cyclone.[/b]
--- Quote End ---
Are your samples coming in at high rates? If that's the case then not only do you need a true FIR chain, but you may need a Stratix device (hardware multiplies). In short if you get the filter into the design, then the rest will be easy.