Can anyone help with RGB color balance?

[astrovienna]

Here's the color data I'm working on, to combine with luminance data:

https://www.dropbox.com/s/09beklcymstoklp/RGBAllMure.xisf?dl=0

That's linear, with Mure Denoise applied to each channel, but nothing else.  I find that after DBE and Photometric Color Calibration, the galaxy has an odd greenish-yellow cast. The blue spiral arms are quite pale.  Any ideas for how to fix it?

This happens now and then with my RGB data.  My guess is that coma and field curvature, which are stronger in my blue data than the other channels, is throwing off color calibration.  Whether that's right or not, I still need to figure out how to fix it.  Any help is appreciated!

Kevin

[topboxman]

Hi Kevin,

I have never seen an image with quite a bit gradients of different colors like yours. I was not able to process it to remove the gradients. It might be best not to do any noise reduction and simply do RGB combination and upload it. Or maybe upload three calibrated, stacked and unprocessed files of Red, Green and Blue XISF images and let us analyze them.

Peter

[oldwexi]

Hi Kevin,
let me underscore Peters comment and let me add also that the flats are not correct.
Its visible with the bright dust circles. This happens only if the flat is to dark.
Maybe you introduced the different colors in the background also with the bad flats.
Also i am surpised about the horizontal banding which is usually only visible in Canon DSLRs.
In addtition i think only the green channel was focused properly and the red one not so good
and the blue one maybe is a little out of focus.

So a lot of improvement has to be done before DBE and ColorCalibration...
if you provide the three channel images seperately also the  masterflat would be interesting
also a raw R and G and B image to see if the many colors came from imaging in the
viennese prater  highlighted by its neonlights....


Gerald

[msmythers]

Kevin

I agree that there are flat issues that need to be resolved. You also have integration issues, the satellite trail. These issues should be looked in to and resolved as it will make processing a lot easier.

I was able with one pass of DBE to remove most of the color gradients. More passes or a more careful application of DBE might help.

I was also able to get a good color balance with Photometric Color Calibration. I used a mask to apply the ColorSaturation tool.

Here is a comparison of the same image. The left is just a weaker brightness/contrast image.


Mike

[msmythers]

Kevin

Just to be more complete here are the setting I used in DBE and PCC. Before running DBE I cropped the edges of the image to remove the noisy overlap areas. I find it's easier to get a better DBE result with things like that removed. I'm also not using the default settings in STF which is why my image looks like it has less contrast. That doesn't change the results just what it looks like on the screen.



Mike

[astrovienna]

Peter, Gerald and Mike, thanks for the suggestions, which I've been tinkering with over the last few days.  Mike, your approach to DBE and PCC is a bit different from mine. I tend to use more points and smaller sampling sizes, but in any case my DBE run did pretty well at removing the gradients, except a tiny bit at upper right.  The key difference I noticed with your processing was your use of region of interest in PCC.  I've just been setting the upper bound to the maximum of the background, but following your approach yielded better results, though I'm still not sure why.  The finished LRGB is here:

http://www.pbase.com/skybox/image/168076264

The new PCC gave a result which was free of the yellowish color cast.  After that I increased saturation in the galaxy itself with a masked CurvesTrans, but I still thought the blue was too weak and did a second CurvesTrans to boost that a bit.  I think that's pretty close to the best I can get with my gear (C11, SXVR-H694, Paramount MX+) and light-polluted site.

As far as flats, I'm not sure there's an easy fix to my issue except for an artificial flat to clean everything up at the end of processing.  I can think of at least three issues.  First, the horizontal banding is probably caused by a reflection off an edge of the OAG pickoff mirror.  That at least should be fixable if I blacken it.  Second, I use a moving mirror (ie, non-Edge) C11, so it's really not possible to get the mirror position for flats to perfectly match the mirror position for each light frame.  I think that explains the faint concentric circular artifacts and the imperfect removal of dust donuts.  Finally, I read a really interesting piece recently that suggested that the color spectrum of the light source for the flat needs to roughly match the sky spectrum, otherwise glows and gradients caused by diffraction and reflections inside the scope won't perfectly match those caused by the sky source.  (I think it was by Richard Wright, on the old SBIG site, but of course now I can't find it.)  I thought that was only an issue, and there only a minor one, with OSC cameras, but if the explanation is right that would explain some of the problem, and also explain why I don't have the same problems in my narrowband images. 

Thanks again for the help.

Kevin

[msmythers]

Kevin

I use the larger wide apart samples and change the downsample value when dealing with slow changing color gradients. A second run of DBE with more precise samples could be run for the normal gradients. Of course don't forget to reset DBE before doing a second run.

As far as the using the region and the upper bounds that is nothing more than the normal use for Background Normalization tool. Using the preview(ROI) you are bounding the tool to only consider those pixels within the ROI. This is why you choose and area void of anything, just background and background you know would be normally neutral. The tool then uses that background model over the entire image. Alejandro Tombolini has some examples of using the upper bound and prviews in the forum and at http://www.pixinsight.com.ar/en/.


Remember default values in tools are only there as a general setting starting point. While they might get you a decent image sometimes they are not optimal. Since image quality can vary greatly those default values can be very wrong sometimes.


Mike

[pfile]

the article you were referring to is probably this: http://diffractionlimited.com/flat-fields-stray-light-amateur-telescopes/

can't you fix the mirror in an SCT and then use an external focuser? i understand that there is not much backfocus available but i think some people do this.

regardless though i think the mirror positioning problem should not result in such bad flattening. you should at least check that your flats were calibrated right. it is certainly possible that you have reflections that are producing the concentric rings, but you should be able to mitigate those by flocking all surfaces that are exposed inside the tube.

rob

[astrovienna]

Hi Rob,

Yes, that was the article I had in mind.  And yes, I need to make sure I'm calibrating the flats properly.  I usually use BPP for calibration, but I need to get deeper into the theory to make sure I'm taking the optimal approach for my gear.

I'd love to have the machining skills (and courage) to machine the mirror cell and lock it, but even if I did I'd run out of backfocus, as you note.  I'm already pushing my luck by being 7-8 mm beyond optimal backfocus.  Good idea about the flocking.  I should get inside the OTA and do some of that.

Mike - I used to do ROIs that way before PCC, when using Background Neutralization and ColorCalibration.  But I somehow thought it was no longer necessary with PCC.  Thanks for the tip.

Kevin

[msmythers]

Kevin

Think of PCC as BN and ColorCalibration rolled into one tool. The difference is color calibration needed the user to roughly identify the white object in the image. With PCC the object or objects(galaxy and stars) are identified through plate solving. The tool can then directly read pixel values and set parameters much like CC. The ROI for the background still needs to be set for the best results. Even with plate solving not all objects(faint stars, dust, gas) will be taken into account so the background could be wrongly interpreted.

One other advantage with PCC is the final image is solved so if your going to be making a mosaic you have one less step to do if using the MosaicByCoordinates method.



Mike   

[stevek]

the article you were referring to is probably this: http://diffractionlimited.com/flat-fields-stray-light-amateur-telescopes/

can't you fix the mirror in an SCT and then use an external focuser? i understand that there is not much backfocus available but i think some people do this.

regardless though i think the mirror positioning problem should not result in such bad flattening. you should at least check that your flats were calibrated right. it is certainly possible that you have reflections that are producing the concentric rings, but you should be able to mitigate those by flocking all surfaces that are exposed inside the tube.

rob

I had issues with flats and reflections inside my scope after the focuser because of the brightness of the flats panel.  Even with already matt black astrophysics adapters and TEC flattener.  I flocked the inside of every component after the focuser with a roll of $8 black velour type flocking material and this fixed it.  The flocking material is absolutely matt black.