Author Topic: Newbie question: Color FITS  (Read 3427 times)

Offline ColinThomas

  • Newcomer
  • Posts: 8
    • View Profile
Newbie question: Color FITS
« on: 2013 September 23 10:22:31 »
Hi,

I am an newbie to ccd imaging, so expect some dumb questions, and have just downloaded the trial version. Basically am saving the pennies for a dedicated deepspace cooled camera next year, and in the interim have just purchased a SBIG STI colour, as both a guidescope (DSLR at present and the deepspace when I've saved up), and as a planetary camera.

I have some images of jupiter and the moon, so want to stack etc. I have found my way to the FFTregistrattion, then use the STF auto stretch. ( I assume this is the correct option for this ?)

One thing is that the color from the image does not seem to be present.  :-[
When I look at the FITs header it is colour-space is grayscale: is this correct ?

I have been exchanging emails with SBIG: their software seems to handle color output from the STI. By they agree that registax has issues (images only look green for jupiter). Is the behaviour from pixinsight implying that SBIG have a generic issue with there FITs color format?

Any pointers to what I should see in pixinsight would be ratefully received, and whether I need any special settings.
This would also be useful longterm in deciding whether to go for a mono or colour deepspace CCD.

Anyway all help gratefully received  :D

best regards

Colin Thomas



Offline pfile

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 4156
    • View Profile
Re: Newbie question: Color FITS
« Reply #1 on: 2013 September 23 11:03:23 »
well for any color camera, at some point in the flow you need to debayer the images. did you do that?

rob

Offline ColinThomas

  • Newcomer
  • Posts: 8
    • View Profile
Re: Newbie question: Color FITS
« Reply #2 on: 2013 September 23 12:03:17 »
Hi,

Thanks for the reply: appreciated
Have been reading in the last hour, and located the debayer option I said I would have dumb questions ;)

Now am running

  • FFTregistrattion
  • STF auto stretch
  • Debayer
  • Dynamic Crop
  • Histogram

and my moon image is looking much better :D

Is this a good recipe then ?

I did try to save as a bmp or jpg direct from pixinsight, but just got black images when viewed in gimp- there were some warnings re 32bit formats - is this to  be expected ? Exporting as fits into gimp to convert to jpeg did work though

best regards

Colin

Offline pfile

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 4156
    • View Profile
Re: Newbie question: Color FITS
« Reply #3 on: 2013 September 23 12:45:19 »
the normal flow is to calibrate, then debayer, then register, stack and then all the other postprocessing.

if you register before you debayer, the debayer results are going to be completely invalid. the debayer algorithms assume that the image is just as it was when it came off the sensor from a 'pixel position' point of view.

jpg out of PI should be okay, so i think what happened is that you just did the STF and did not actually apply it to the image. STF is only for viewing linear images on screen and does not actually touch the data. if you like the autostretch that STF computed, you can copy it to the HistogramTransformation process by dragging STF's triangle to the bottom bar of the HistogramTransformation process (the horizontal strip where HT's triangle lives). then you can drag HT's triangle to your image to apply the stretch for real. don't forget to turn off the STF after that's done or your image will more than likely be completely white (but again that's just the screen transfer function, the data underneath is not all whited out...)

rob

Offline pfile

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 4156
    • View Profile
Re: Newbie question: Color FITS
« Reply #4 on: 2013 September 23 13:31:14 »


I have been exchanging emails with SBIG: their software seems to handle color output from the STI. By they agree that registax has issues (images only look green for jupiter). Is the behaviour from pixinsight implying that SBIG have a generic issue with there FITs color format?


i have not used registax with raw images but i'd be surprised if registax did not know how to debayer the popular bayer patterns. i don't think there are any generic issues with color fits other than the below "gotcha"

one very important thing that i did not mention is that fits files can be read as top-down or bottom-up, meaning, the XY position of the 0th pixel in the file can either be at the top or bottom of the frame. since the debayer algorithms go by "dead reckoning" if your fits reading direction does not match the direction that the manufacturer is writing out the file, then your colors will look crazy after debayering. IIRC you can choose the "opposite" debayer pattern (BGGR instead of RGGB for instance) to get around this. as i've gotten older though i'm getting worse at spacial stuff so that might be wrong.

at any rate, one way to unambiguously determine the right settings is to take a terrestrial daylight picture with the camera (if possible) and then experiment with the debayer settings and fits reader direction until you have the right one. even on an astrophoto it's usually obvious that you have the wrong one (screendoor effects left in debayered image, or just crazy colors)

note that a green cast is sometimes not indicative of a problem. since the human eye is most sensitive to green light, bayer matrices have 2 green pixels. a lot of times right out of the camera the green channel will dominate because of this. if you unlink the channels in STF (the chain link icon at the top) STF will normalize the channels independently and fix that up for you at least as far as an initial view of the linear data goes.

rob

Offline ColinThomas

  • Newcomer
  • Posts: 8
    • View Profile
Re: Newbie question: Color FITS
« Reply #5 on: 2013 September 24 11:55:55 »
Hi Rob

Thanks for your replies: appreciated.

I have emailed SBG for the exact debayer setting for the SI colour I have, and am awaiting a reply.

In the interim, have located the debayer batch
Script->Batch Processing->Batch debayer

So for my odd 100 Moon fits, pass through this and o/p files

Then go to:
Scripts->Utilities->FFT Registration

and o/p files again to a directory

Then take these files into ImageIntegration

Then play with the HistogramTransformation.


What other tools are then worth looking at for lunar surfaces ? Any pointers welcomes  :D

best regards

Colin

Offline pfile

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 4156
    • View Profile
Re: Newbie question: Color FITS
« Reply #6 on: 2013 September 24 12:47:51 »
Deconvolution is a pretty important tool for planetary images. check this tutorial out:

http://pixinsight.com/examples/deconvolution/moon/en.html

very technical but accessible i think.

for deconvolution to have physical/mathematical meaning it should be done on linear (unstretched) images. of course you will find people doing deconvolution on nonlinear images, which does of course have a sharpening effect, but it's not "proper".

generally speaking, people usually extract the L* from color lunar images and process that, but it is possible to work with color moon pictures. apparently the moon does have some subtle coloration but when starting out it might be best to stick with monochrome images.

once the lunar image has been deconvolved and stretched, i find that HDRMultiscaleTransform (thru a mask), LocalContrastEnhancement (again thru a mask) and finally some curves for added contrast work well.

rob

Offline Phil Leigh

  • PixInsight Addict
  • ***
  • Posts: 220
    • View Profile
Re: Newbie question: Color FITS
« Reply #7 on: 2013 September 26 02:25:26 »
The moon certainly has rather a lot of colour but you will need to play with the saturation curves to bring it out! Of course some people prefer a grayscale moon - it's an aesthetic choice.

You have to get the background neutralisation and colour calibration right before boosting the saturation. You need to keep an eye on what is happening to the luminance data - it is very easy to end up with a blue or orange moon :-)