I'm a bit confused about how filtering an image based on frequency could actually be done in practice — this isn't a simple blur, right? (The frequency domain visualizations of the blurred images were much less clean than this.) Intuitively, it seems like we're losing information going from the spatial domain to the frequency domain, so we can't just convert back after modifying the frequency content of the image. What does each point in the frequency domain correspond to in the spatial domain/in image terms, and how would you actually perform this sort of filtering?
I'm a bit confused about how filtering an image based on frequency could actually be done in practice — this isn't a simple blur, right? (The frequency domain visualizations of the blurred images were much less clean than this.) Intuitively, it seems like we're losing information going from the spatial domain to the frequency domain, so we can't just convert back after modifying the frequency content of the image. What does each point in the frequency domain correspond to in the spatial domain/in image terms, and how would you actually perform this sort of filtering?