The claim that vertical and horizontal lines in the transform's output are caused by natural tendency to have vertical and horizontal edges seems doubtful. I su

The claim that vertical and horizontal lines in the transform's output are caused by natural tendency to have vertical and horizontal edges seems doubtful. I suspect they are caused by discontinuities between the left and right image borders. DFT is nearly unchanged (only phases change) under cyclic shifts of the input. Thus a discontinuity inside the image is "just as bad" as a discontinuity between the left and right border of the image. If you want to DFT an image for analysis purposes (ie. if you don't need the transform to be invertible), you can use the trick from the "Periodic plus Smooth Image Decomposition"[1] paper that extracts a "periodic approximation" of the image. It's used e.g. for DFT-based texture enlargement/periodization. [1] https://hal.archives-ouvertes.fr/hal-00388020v1/document https://hal.archives-ouvertes.fr/hal-00388020v1/document