Aside from the discussion on whether you need isolation, these are the possible techniques that you could use. I hope I haven't omitted any.

I can't think of any instance where you wouldn't need an isolated power supply as well to power the circuitry on the input side.

Linear Optical
--------------
I am aware of two products. There is the LOC110 from Clare (which has a number of second sources (with different numbering but identical specs) as I recall including Siemens (or whatever its spin off is called.)) The second is from Avago- HCNR201. I have worked with both with very good results. Both require op-amp circuitry around them to achieve linearity performance.

Linear Magnetic
-----------------
You could convert the signal to pulses and transmit this across a transformer, but this is probably better left to a commercial Isolation Amplifier rather than roll your own

Digital Opto-
------------
You can use digital isolation as suggested by Erik, where you convert the analog signal to a frequency with a VFC (see Analog Device, Linear Tech or National) and then cross the isolation barrier with a digital opto. One advantage is that you don't need an A/D on the micro and can possibly achieve much better than 8 bit resolution.

You could use a 2 micro approach with one on either side of the barrier, or a serial A/D on the input side communicating with the micro on the output side through digital optos.

Isolation Amplifier (as a complete block 1 or 2 ICs)
-----------------------------------------
Analog Devices, and TI (Burr Brown used to make) which normally works on magnetic or optical principles.

Burr Brown make isolation amplifiers that worked on capacitive coupling, ISO124 amongst others.

Relay Isolation
---------------
A very old technique was a "flying capacitor" connected to the common terminals of a 2 pole changeover reed relay. The capacitor is connected to the input voltage and charged and then switched to connect to the output circuitry- interesting from an historical perspective.