# Relation between Gaussian and Savitsky-Golay kernels

I would like to make a mathematical comparison between two methods in my field. One smooths with a Gaussian filter, and one smooths using a second-order Savitsky-Golay filter. Let's say the window size is the same in each case. Is there any way to formally compare these two approaches in terms of their impact on a function f? Can one be expressed in terms of the other, or in some common framework?

• Well, you could compare the frequency transfer function, or the impulse response function. Probably the best basis for comparison should depend on the function f. – Dave Kielpinski Feb 26 '15 at 5:33
• Great. I was thinking after posting this that a DTFT approach might be fruitful, sounds like that is worth trying. – barnhillec Feb 26 '15 at 11:02