Discrete Fourier Transform Questions

Hi Folks, I am currently trying to analyze the “high” frequency content of a Fourier transform output using the following documentation as a guide:

docs.opencv.org/2.4/doc/tutorials/core/discrete_fourier_transform/discrete_fourier_transform.html

I have the following questions which I hope someone can help me with.

1) To analyze the frequency content, is it appropriate to analyze just the magnitude data (or should I also consider some other aspect of the DFT)? My goal is to check how good the focus of an image is (since high frequency content in the DFT is an indicator of a sharp image).

2) I believe that the further you move from the center of the “centered/adjusted” DFT image, the higher the frequency content so I intend to analyze a band/region that is a certain radial distance from the center. I might take the average frequency magnitude in that region to get some sort of metric. Would this be a “fair” or standard way of analyzing a DFT image for high frequency content?

3) I understand that the Fourier transform output for a real image has symmetry i.e. quadrant 1 has the same frequency components as quadrant 3 etc. so is half of the DFT image redundant? I.e. I just need to analyze the top or bottom half of the DFT image?

I hope these questions make sense and that someone might have some experience in this area.

Thanks, T4o.

edit retag close merge delete

Sort by » oldest newest most voted

1) "To analyze the frequency content, is it appropriate to analyze just the magnitude data" read this http://www.comm.utoronto.ca/~dkundur/... and read this too. IMHO answer is no : phase and magnitude are important your question is z= r exp(i phi). Is r most important than phi ?

2) It will depend of what you want : use google key words steerable filters

3) image is made of real number : bacsic properties are true for 2d signal

more