# Discrete fourier transform and Frobenius norm

I have thousands of data files which are essential DFT data of plots such as this:

And a DFT of the plot with a "Hard" threshold of 0.9 gives me:

This DFT is just the left top corner of the DFT data with the DC value removed.

The reason I used a threshold value of 0.9 is because in the DFT coefficient matrix that I plotted above, most of the values of very small (~10^-4) and there are maybe 1-2 values that influence the DFT and they are above 0.9

However, it was suggested that I use a Frobenius Norm of my DFT data to threshold my plots.

Why is that? I can't use a Frob. norm as a threshold because it is a very large value ~100 and my DFT coefficient matrix is populated with numbers <100 generally

Should I just divide my DFT data with the Frob. norm? I have read the following posts 1 2 3

~Concerned Mechanical engineer

-
How are you defining the Frobenius norm? –  Manos Oct 13 '12 at 17:38
I am not. Mathematica is. I just plan on using the Frob Norm for my work. –  drN Oct 13 '12 at 17:38