According to Zekun, this part can be tricky.

I think the coefficients z_k should be calculated for k = 0, 1, 2, 3, ..., N-1

add more override and static methods for mathematical operations.

Now using in src/

javadoc -private -d ../doc/ -J-Duser.language=en_US -encoding UTF-8 -charset UTF-8 -docencoding UTF-8 -subpackages fourier

I want to add this functionality to it, so that we can recall coefficients calculated.

Now we can draw several points on the panel and get the corresponding curve. But due to our poor algorithm the Array of sample points and the coefficients now has to have the same size. Which means we have to set the slider n to equal to the number of samples N.

Please read the comments around 370L in GraphicFitting.java

If we sort the sample points by their angles, thus form a clock-wised order, the absolute value of coefficients for high frequency items could be reduced.

Can you try fix that?

MATH IS VERY COMPLICATED.

I wrote some code mainly implements the Discrete Fourier transform algorithm. N sample complex numbers are accepted. Another N complex number calculated.

after an image was loaded to the panel, if the whole window is manually resized, the image would not move to the center of the panel immediately. Maybe one more repaint operation is needed.

It seems the performance bottleneck is from the curve plotting.

To update the coefficients table that is bound with the array coef[], add the following code snippet

for (int i = 0; i < n; i++) {
  coefficients_table.setValueAt(coef[i].re(), i, 1);
  coefficients_table.setValueAt(coef[i].im(), i, 2);
  coefficients_table.setValueAt(coef[i], i, 3);
}

which is already there at 353L of GraphicFitting.java

according to https://help.github.com/articles/creating-project-pages-manually/

in branch gh-pages

view at http://zvant.github.io/FourierTransformFit/

javadoc command:
javadoc -private -d ../../Pages/FourierTransformFit/ -J-Duser.language=en_US -encoding UTF-8 -charset UTF-8 -docencoding UTF-8 -subpackages fourier

Please leave related comments in this issue.

In discreate FT, N-1 -th order is actually equivalent to -1-th order. By folding the higher N/2 orders to negative orders, the high order coefficients are much lower and the resulting curve is smooth.