Can this FFT be standalone? I have PSCS 5.1
Please see the tutorial in the link below on how to get rid of 'edge banding' effects on some images during restoration.
https://www.photoshopgurus.com/forum/general-photoshop-board/56261-fft-plugin-macs-dj-joofa-being-made-available.html#post1533724853 (https://www.photoshopgurus.com/forum/general-photoshop-board/56261-fft-plugin-macs-dj-joofa-being-made-available.html#post1533724853)
BTW, the '_C' version of the filters don't exhibit visually noticeable edge banding and produce cleaner edges. The procedure in the above link is not needed with them.
Joofa
You can use ImageJ (Windows, Mac) for this purpose. Gimp can also be used with some free stuff.
are there tutorials for more general applications of FFT?
I know it is sometimes used to demonstrate telltale signs of noise reduction.
Can this FFT be standalone? I have PSCS 5.1
.....
Yes, I have it, but are you going to make yours standalone?
How does one use these in ImageJ? I downloaded it, put the two plug-ins into the plugins folder, but they don't show up in ImageJ's Plugins menu. Choosing one of the FFT plugins with Plugins > Install gives the error message, "File name must end in '.txt', '.ijm', '.js', '.bsh', '.c|ass', '.jar', '.zip', '.java' or '.py'" and Plugins > Compile and Run gives the error message, "File name must end with '.java' or '.cIass'."
How would you tell your grandmother to get these to work?
Oh, and so it does in ImageJ 1.4.9.... I asked because you said in an earlier post that your plug-ins work in ImageJ and GIMP, so I just downloaded ImageJ tonight solely to try them because I'm in Photoshop CC (2013).... But don't your plug-ins have that edge enhancement that ImageJ's native ones don't? (And I'm still not clear if your current ones have that edge enhancement yet, or whether you're still waiting for enough donations.)... And doesn't ImageJ's FFTs need a fixed pixels aspect ratio crop, whereas yours will work with any aspect ratio?
(1) SSFTIt might be interesting to compare spatially windowed FFT analysis to the spatially windowed DCTs used in e.g. JPEG compression. Oh, and why does the audio people use (almost exclusively) MDCT (50% overlap _and_ critically sampled) for compression purposes, while the imaging community does not?
(2) VideoI am guessing that you are doing a (separable) 2-d FFT of each frame, animated in time? It might be interesting to do a 3-d FFT of such a sequence. Not that my skills at manual interpreting 3-d FFTs are any better than my skills at manually interpreting 2-d ones...
For retouching gurus, please see the following link for free Mac Photoshop FFT / IFFT 64-bit Plugins for CC 2015 (OS 10.9+).
http://www.retouchpro.com/forums/software/38969-developing-mac-photoshop-64-bit-fft-ifft-plugins.html (http://www.retouchpro.com/forums/software/38969-developing-mac-photoshop-64-bit-fft-ifft-plugins.html)
Sincerely,
Joofa
I believe I'm the only Lundberg02 online and I have engaged with some of the greats because I am fearless in my ignorance, but I wish I knew what specific discourse inspired your kind offer.
I worked with the FFT almost as soon as Cooley and Tukey published because it was essential to anti submarine warfare sonar signal processing. The company I worked at the time built the first 16 bit mini to do it. We used radix four to simplify. I was amazed that an experienced operator could give you the hull number and the captain's name from that slightly wiggly line crawling down the chart. I believe that Cooley and Tukey probably developed the FFT for use with the SOSUS arrays that still defend our coasts.
It might be interesting to compare spatially windowed FFT analysis to the spatially windowed DCTs used in e.g. JPEG compression.
Oh, and why does the audio people use (almost exclusively) MDCT (50% overlap _and_ critically sampled) for compression purposes, while the imaging community does not?
I am guessing that you are doing a (separable) 2-d FFT of each frame, animated in time? It might be interesting to do a 3-d FFT of such a sequence. Not that my skills at manual interpreting 3-d FFTs are any better than my skills at manually interpreting 2-d ones...
Hello,
I am certainly new to this "plugin/filter" and I'm definitely interested in using it for some of my retouching jobs. I have downloaded the CC 2015 version, which also seems to be compatible with OS 10.9 according to the above information. Is that indeed the case? Otherwise I'll have to wait until I decide to subscribe. Verification would be much appreciated, either it will or will not work with OS 10.9+.
Thank you,
Gary
Yep, that is what I have been thinking of lately also. I shall make a DCT plugin for Photoshop and compare it to FFT.Note that the DCT can be calculated provided that you allready have a FFT implementation using some pre/post-processing steps.
I would think that the DCT might give a better response regarding image edge / boundary effects.Yes, the assumption of DCT (or at least the popular kind) is that the signal is "mirrored" at each block boundary. Clearly, this is an assumption that better suits images (as evidenced by its usage in image scaling applications) than the FFT assumption that the segment is periodically repeated.
And, the DCT representation of signal may be more compactified than DFT.I have seen this attributed to the same edge assumptions. I think theoretically, the KL transform (or, equivalently, PCA) will find the most energy compacting linear transform of a given dataset. Textbooks typically claim that the DCT is a practical close approximation to the KLT for many real-world signals without having to adapt the transform to a particular signal, transmit the transform (adding bandwidth) and dealing with generic (slow) code.
Perhaps, audio signal are more nonstationary, and might require so. Also, images usually have a much larger amount of data points so perhaps people don't want to increase the number of analysis points.I think the latter. Doing 50% overlap in a 2-d signal would mean 4x the datapoints processed? Also, I am guessing that small edge artifacts are more acceptable visually than audibly (for sound, you might get some audible click at a fixed rate of e.g. 10 milliseconds).
Yes 2-D FFT of separate frames in time. In computer vision they call such processing 2.5D! ;D One can take a singular FFT along the time domain of these FFT'd image sequence to get a full 3D FFT.The question is if we can get any further insight from transform along the 3rd dimension as well.
Doing 50% overlap in a 2-d signal would mean 4x the datapoints processed?
Also, I am guessing that small edge artifacts are more acceptable visually than audibly (for sound, you might get some audible click at a fixed rate of e.g. 10 milliseconds).
The question is if we can get any further insight from transform along the 3rd dimension as well.
I think in audio spectral estimation after dividing a signal into overlapping segment, each segment is windowed also. May be that is also a reason for overlap, to make those data points that are windowed out (or lowered in value) have a contribution to spectral estimation.In the case of the MDCT, you have 50% overlap but still critical sampling, i.e. the transform produces 512 spectral samples out of 512 temporal samples. The trick is to choose a window with particular properties (timedomain aliasing cancellation) and to depend on the overlapping analysis/synthesis.