Introduction to Fourier Transforms and Quantitative image Processing

Hey Scientists,

For those looking for a beginner level introduction to Fourier transforms and quantitative image processing, here is a talk I gave this year at the Advanced Imaging Methods Workshop:

It starts with an introduction to 2D Fourier Transform power spectrums and how to interpret them, then goes into filters and what each filter does to the information content of the image. I then go through real-world examples of how you can use just median filters, Gaussian blurs, and image calculator to denoise images and movies that appear at first glance to have an intractable background.

I then cover the basic math underlying deconvolutions, how perfect deconvolutions are theoretically possible, and then why real-world deconvolutions are inherently imperfect. And finish with using custom kernels to enhance image contrast (specifically showing the emboss kernel).

Unfortunately, this was a talk given virtually, and the platform was having a pretty significant video lag. That said, you can follow along with the talk yourself, go through the macro demo, and get all the example images covered in the talk here:

Ben Smith


Great talk! I think one of the issues was the compression level they were using for the video stream - in some cases you can really see the encoding - and in others the little spots don’t show up at all. Nothing you had control over though.
I often wish I could see my image analysis talks from the viewer’s point of view - during the talk - so that I know if what I am seeing is the same as what they are seeing!!

1 Like

I can see the Matrix…

1 Like

@allopatry If the universe is a hologram, does that mean everything we perceive is just the inverse transform of reality?


A hologram per se has nothing to do with the Fourier-Transformation.
Only some very special holograms make use of the FT.

And never forget:
Holography is wavefront reconstruction that says nothing about the inside of objects and at best something about the hull.