I think Nagarajan already told that the range of data that need to be linearised is very large to fit in a LUT.

I once used a reduced lookuptable of the following form
a1,b1,c1
a2,b2,c2
....

if input = a1, output is = b1
if input >a1 <a2 output is b1 + ((input - a1) * c1)

it is, of course, slower since you need to search for tha 'a' rather than 'just lookup' but the compresion ratio for the app where I used the technique was fantastic.

Erik