PixInsight 1.7.0 codename Starbuck released

[RBA]

If you are a Windows user, remember that you should uninstall the current version (Start > All Programs > PixInsight Platform > Uninstall) before installing 1.7.0.

Unless I missed something, additional modules/scripts that I had installed on my previous install are gone after the uninstall.

It'd be nice if the uninstall program would warn you when it's about to do that, and maybe give you the option, perhaps with a warning note that older modules may not work or whatever if that's a concern.  Yes?

[Ioannis Ioannou]

FourierTransform and InverseFourierTransform allow you to perform discrete Fourier transforms and manipulate the transform components directly (either as the power spectrum -magnitude and phase- or as the real and imaginary components). Then you can apply the inverse transform from the modified components. An extremely interesting image analysis and processing tool in the frequency domain, conceived by PTeam member Carlos Milovic.

Muchas gracias señores (although my current camera has minimal fixed pattern and banding noise, still there are many other ways to use this). How about some information how to use it ?
I used some old bias that I know that they had fixed pattern noise, and tried a bit of clonestamp on the magnitude DFT (a bit randomly I'm afraid). Then I used the reversed transformation. Which of the DFT components should be the first and which the second? The results I got were not exactly the expected but for sure were interesting.

[Carlos Milovic]

Hi John
I plan to write some specific filters for fourier domain images, like band-pass, notch, etc. Also, the module may benefit from further processing, like using user defined zero padding (or mirror padding), etc.
About documentation... formal tutorials are not in the near future... If you want to ask something more or less specific, I'd be happy to help/anwer, but I'm afraid that I have not the time to sit down and write something better.

[Ioannis Ioannou]

No problem Carlos. My initial intention was to use it as a fast visual tool to evaluate the quality of the camera, the same way Stark has shown with ImageJ on his articles. At this point your implementation IMHO is superior to the one of ImageJ or Iris.

Then I was interested on using it to correct banding and fixed pattern noise on my previous camera, the same way as Iris is doing it with the FFTD, FFILL and FFTI commands (http://www.astrosurf.com/buil/iris/new410/new410_us.htm at the bottom). Since I have no theoretical background (I understand the concept and I suspect the possibilities, bit I have no real knowledge) and my camera now does not have this problem, I'll just wait for your filters (which I suspect will be very very interesting). Meanwhile I used them a bit and I was intrigued.

[Ioannis Ioannou]

Hey, great job !!

Among the others, I was very impressed with the speed improvements and resource management.
Usually with the previous versions, I had to run Process Lasso on my laptop when I was running PI with something heavy (like StarAlignment) while I was doing other things (this is a job laptop). Process Lasso used to put 1.6 under lower priority allowing me to continue. But with 1.7 this was not needed !! I tested with a special heavy case with many frames and it did it pretty fast and with the CPU usage well under 60% (it was 100% with 4 cores when at 1.6)

Well done !

[Carlos Milovic]

John, from what I saw form Iris' documentation, this seems to be a simple implementation of a notch filter. As you figured out, a straighforward way to do that right now is using the clone stamp. Just remember that those points (that indicate a "defined" sinosoidal pattern) always comes in pairs, opposite in both axis in relation to the centered magnitude spectrum.
What I want to implement is a sort of adaptative notch filter, that finds a weighting function, which supress the "banding noise" while preserving the information beneath it (in the fourier space). I may end doing a dynamic process... since I suspect that this still will have a lot of trial and error.

[Jordi Gallego]

Thanks Juan for the hard work on that new release. Installed and working OK on win 64. The project funcionality is really wonderfull, congratulations!

[Ioannis Ioannou]

What I want to implement is a sort of adaptative notch filter, that finds a weighting function, which supress the "banding noise" while preserving the information beneath it (in the fourier space). I may end doing a dynamic process... since
I suspect that this still will have a lot of trial and error.

I'm pretty sure that there are a lot possibilities beyond the ones I had in my mind. And these tools will be very helpful in cases that we may not thought about yet. Looking forward to.

[RBA]

Unless I missed something, additional modules/scripts that I had installed on my previous install are gone after the uninstall.

It'd be nice if the uninstall program would warn you when it's about to do that, and maybe give you the option, perhaps with a warning note that older modules may not work or whatever if that's a concern.  Yes?

Maybe something to contemplate in the future?