It seems strange that a single piece of glass can compute the Fourier transform of an image, but it is true (sort of). I explore an optical arrangement known as a 4F correlator, and note the practical considerations that are rarely discussed in existing sources.
I implemented an image feature detector using accelerated vector processing using fft a while back and the "filter" (i.e. the part of the image you're searching for) had to be rotated 180 degrees when it is combined with the input image. Perhaps that'll help you get a more clear result?ReplyDelete
Also, keep in mind that the result of a feature detector is a point (or points) of maximum intensity indicating the location in the original image. I'm not sure how clear this point will be given the extra phase information in the result. Awesome and interesting stuff, thanks Ben!
I just stumbled onto your blog tonight. You are seriously my hero. I wish I knew half as much as you. Keep it up and thank you. I am studying CNC and electronic systems in school, but they don't move half as fast as I would like, so 90% of my education is online. Your videos are super inspiring.ReplyDelete
Very interesting. This inspired me build a simplified 4f correlator that projects a grating pattern onto the wall. I have also done a review of how a lens can be used to perform the Fourier transform that might explain how it works.ReplyDelete