Posted by
Michael Schmid on
May 03, 2015; 1:35pm
URL: http://imagej.273.s1.nabble.com/frequency-domain-filtering-using-a-mask-tp5012677p5012686.html
Hi Mike,
the Fourier transform is a linear operation, so instead of taking the
difference of two FFTs, you can also take the difference of two images and
then to the FFT. Use 32-bit data, i.e. real numbers, so you can have
negative values.
For finding the difference of two meshes, I would try to align them first,
either by finding the maximum of the correlation or, if there are
distortions by some plugin like bUnwarpJ (I tried a sample image with all
weights set to 10), then subtract the registered images.
An alternative might be taking the FFT of the reference grid, thresholding
it to get all non-negligible Fourier components, inverting it, and using
it as a Custom Filter. This will make the defect stand out, but it won't
work very well near the edges.
Michael
_______________________________________________________
On Sat, May 2, 2015 21:17, Mike Morrione wrote:
> New to imageJ
>
> I have two images. First image is a uniform mesh and the 2nd image is the
> same mesh with a defect ( a small round dot). Images are black and white.
>
> I am experimenting with finding a defect based on a reference image (image
> without defects). I believe I should be able to take the difference of the
> FFT's of each and use the result to locate the defect after inverse
> transform.
>
> However after taking the difference using image calculator the result is
> no longer in the frequency domain so I can not do the inverse transform.
>
> I would appreciate any help.
>
> Thank you.
>
> Mike
>
> --
> ImageJ mailing list:
http://imagej.nih.gov/ij/list.html>
--
ImageJ mailing list:
http://imagej.nih.gov/ij/list.html