#45 closed defect (fixed)
psImageFFT to return half results
| Reported by: | Owned by: | eugene | |
|---|---|---|---|
| Priority: | high | Milestone: | |
| Component: | PSLib SDRS | Version: | unspecified |
| Severity: | normal | Keywords: | |
| Cc: |
Description
Just need to double-check an assumption I made:
Since FFTs of real data results in symetry, is it acceptable that psImageFFT
return an image (nx/2+1) by ny pixels (as FFTW does), or should I really reflect
the resulting image to return a nx by ny pixel array to match the input image's
dimensions?
(nx and ny are the size of the input image in the x and y dimensions, respectively)
Change History (2)
comment:1 by , 22 years ago
comment:2 by , 22 years ago
| Resolution: | → fixed |
|---|---|
| Status: | new → closed |
As per meeting this morning, added to the SDRS:
"The output from a FFT and power spectrum shall be of the same size as
the input. In the future, if this adversely affects performance, this
will be revised so that the redundant information will be neglected."
Note:
See TracTickets
for help on using tickets.
This is causing abit of a problem for me when implementing the *PowerSpectrum
functions, as I don't really know how to interpret the fourier domain data.
If r->c FFTs were only supported, I'd just assume a half frame, but a c->c FFT,
of course, must return a full frame fourier image.
One option is to remove complex to complex forward FFT support (so I can assume
a complex fourier that is nx/2+1 in size of the original).
Another option is to pass in a flag to the function indicating how I should
interpret the data.
(FYI: I just implemented cabs2/N2 to each DFT element for now, but that
doesn't match the ADD and assumes a r->c transform.)