Author Topic: Need help with color calibration  (Read 5469 times)

Offline erikgu

  • PixInsight Enthusiast
  • **
  • Posts: 94
    • View Profile
Need help with color calibration
« on: 2012 May 27 15:38:41 »
Hi !
This is the second time i process this image of M42. The first time the stacking was done by use of DeepSkyStacker, this time its is doen by the excellent BatchPreprocessing  script.
I think the PI script gives a much better result. My problem is to get a good result of the ColorCalibration, few stars and little background. Have done a few tries but i am not satisfied with the result. So it would be nice if any could help me.
 
The image is a 64 bits HDRComposed image consisting of 5,3,240 and 600 sec. exposures.
http://dl.dropbox.com/u/59585830/Files/HDR2.tif

Erik G

Offline georg.viehoever

  • PTeam Member
  • PixInsight Jedi Master
  • ******
  • Posts: 2131
    • View Profile
Re: Need help with color calibration
« Reply #1 on: 2012 May 28 13:10:41 »
I am unsure what the problem is. Can you show screenshots of your unsatisfying results? I can get the stars to be white very easily (only modified Background upper limit to 0.0002). The color effect of the nebula itself is more of an artistic question and can be modified for instance with ColorSaturation and other tools. Also see http://pixinsight.com/forum/index.php?topic=4308.msg30325#msg30325 .
Georg
Georg (6 inch Newton, unmodified Canon EOS40D+80D, unguided EQ5 mount)

Offline erikgu

  • PixInsight Enthusiast
  • **
  • Posts: 94
    • View Profile
Re: Need help with color calibration
« Reply #2 on: 2012 May 28 23:25:50 »
Hi and thanks for helping George.

Hope this shows what i mean. The ColorCalibration on this image was done by aggregting previews with some of the "surrounding" stars as white reference and a small preview in the lower left corner as dark reference. The image has, atleast from my point of view, been very red. Am i doing the CC wrong when i select the surrounding stars as white reference ?





Erik G

Offline Juan Conejero

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 6934
    • View Profile
    • http://pixinsight.com/
Re: Need help with color calibration
« Reply #3 on: 2012 May 29 06:33:25 »
Hi Erik,

Let's see... you have H-alpha emission (red), reflection nebulae (blue), lots of O-III (teal) and H-beta emission (cyan), and lots of stars (multicolor) in this image. Your WHOLE image is a fantastic white reference. Sounds crazy? Let me demonstrate it.

First step: Neutralize the background with the BackgroundNeutralization tool. This is the image before background neutralization:


and this is after background neutralization:


Note that I have defined a preview over a relatively free background region (actually, nothing is 'background' in this region of the sky, but considering the depth of your image the selected area is a good aproximation). This preview has then been used on the BackgroundNeutralization tool to define a region of interest. The rest are default parameters.

Once the background is neutral, we can apply the ColorCalibration tool. The same region of interest has been selected as the background reference.


Note that I have just disabled the structure detection option. The rest are default parameters. This effectively has selected the whole image as white reference. The result is quite good, as the next steps will show more clearly.

Now the next step is a nonlinear stretch with HistogramTransformation. This process is extremely critical for high dynamic range images like this one. This is the linear image without STF visualization:


and this is the stretched image:


Since this is an HDR image the next step is obvious: compress the dynamic range. HDRMultiscaleTransform is the tool of choice for this task in PixInsight. This is before HDRMT:


and after HDRMT with 6 wavelet layers:


The image has light pollution gradients and/or residual vignetting, which complicate further processing (irregular illumination problems should be fixed on each individual HDR component before HDR composition). However, I have cropped the central region and tried out a color saturation curve with a luminance mask, just to prove the nice color balance we have achieved:



This image is a nice example of our deep sky color philosophy. There is no such thing as 'real color' in deep sky astrophotography. Our color calibration tools and techniques have been designed to maximize information representation through color, without favoring any particular type of object as a white reference (based on the properties of the human vision or other criteria). By using the sum of light from the whole image as a white reference, we have maximized the visual separation between the different light sources in the image. Since this image provides a rich sample of light throughout the whole spectrum, the achieved white balance is unbiased. This strategy has been conceptually similar to our 'standard' technique of using a nearby spiral galaxy as a white reference: the integrated light of a spiral galaxy provides a huge sample of all existing spectral types and DSOs, which makes it an excellent unbiased white reference.
Juan Conejero
PixInsight Development Team
http://pixinsight.com/

Offline oldwexi

  • PixInsight Guru
  • ****
  • Posts: 623
    • View Profile
    • Astronomy Pages G.W.
Re: Need help with color calibration
« Reply #4 on: 2012 May 29 07:31:37 »
Thanks Juan
for this image processing example.
It raised another question for me.
In the curves window you used "c" instead of "S" for increasing saturation.
What is your advice? When to use "c" and when "S" in the curves process?

Aloha
Gerald

Offline erikgu

  • PixInsight Enthusiast
  • **
  • Posts: 94
    • View Profile
Re: Need help with color calibration
« Reply #5 on: 2012 May 29 14:57:55 »
Hi and thanks Juan.

This was an excellent and helpful "tutorial". I have struggled with CC on this image for a long time and had no idea that the image itself could be used as white reference.  I have to try to get rid of vignetting on the individual images and compose them again to hopefully get a better result. Thanks again.

Erik G