2-Dimensional Discrete-Space Fourier Transform

2D discrete-space Fourier transform, the convolution-multiplication property, discrete-space sinusoids, 2D DFT, 2D circular convolution, and fast computation of the 2D DFT.

[Stucky]

Wow, thank you for that. I'm doing an internship with radio interferometers and this helped me understand the stuff tremendously. Thank you!

[harshjain]

Great. Fftshift was what I was missing. Great explanantion.

[cooper]

Thank you so much for making this video!

[brevin-tilmon]

DSP final exam coming up. Thank you for taking time to make this video.

[jcgongavoe]

I had a stroke when he said v is the horizontal frequency...

[allistermason]

You are a fantastic human begin. Thank you.

[vikaschoudhary]

You are awesome !!! Thank-you very much for the easy explanation.

[Daniel-enbt]

You are great! Now I understand WAY better!!!! 

[johnsonkuan]

Thank you so much for this great lecture!

[pantich]

Great video and great explanation. I'm new to this topic and tried to visualizza the Fourier Transform of the photographer like you show at min 5:45. I've tried with Octave, Scikit image and Numpy, using fft, fft2 and fftn. Couldn't get the same result, mine doesn't get even close to that. I'm transforming the image into a float skimage.image_as_float(), then calculating the fft and printing it with mathplotlib.pylab with
Am I doing something wrong? Thanks

[luismolina]

Barry, If I have a picture of Fourier Plane, How I can get a picture original?

[ShankhaAcharya]

Excellent Video
I had confusion what Fourier Transform to do in Image domain?
how image pixels forms wave form?

those fundamentals are cleared and got little more than that
Now I'll go back to my book and revise in the light of your explanation
thanks a lot :D

[paraescucharrap]

Bro, thnks so much! this helped me a lot with the Uni!

[antonbursch]

excellent videos!  i love them!  and i have a question.

i am using this to convolve an 80x120 image with a 30x30 kernel.  it's very fast compared to the slower pixel by pixel approach, but i tend to get over saturated images.  for example, 3/4 of a convolved image is pure white.

i suspect that that this is a scaling issue, because it would happen using the slow convolve approach unless i divided the sum of the convolved pixel by the width*height of the kernel i used.

i am trying to figure out how to scale down the pixels when using dft to convolve an image.  

any ideas?

[mansukhkaur]

What does - sign signifies in the formula

[NHK]

Could you please tell me what does the value F(u, v) corresponds to in frequency domain? The amplitude?

[vikaschoudhary]

You are awesome !!! Thank-you very much for the easy explanation.