Loss of colour on mosaic construction

[Andre Vilhena]

Hello everyone,

I am creating a mosaic of Rho Ophiucus, one panel whose main features are the dark and reflection nebulas (top - P1) and the other whose main features are the orange and red emission nebulas (plus M4) (botton - P2) and I have noticed that on the bottom of the mosaic, the red hues are mostly lost, albeit they're present in the corresponding panel (P2).
For comparison, I have made a similar stretch and colour saturation on the pre-merged frame and on the same area of the mosaic (I created a preview of the area and separated it from the mosaic) - please see belo photo "PanelComparison", where on the left is the pre-merged frame and on the right the preview.

The frames were processed the following way: Crop, DBE, RGBWorkingSpace (1, 1, 1), MMT for noise reduction with inverted low contrast luminance mask, HistogramTransformation (slight stretch on each until the histograms match). Then, I used GradientMergeMosaic (Average) after registration to merge them.
Below I also attached the following histograms:

P1_Histogram - histogram of top panel immediately before merging
P2_Histogram - histogram of bottom panel immediately before merging
BottomMosaic_Histogram - histogram of the merged mosaic immediately after merge (of the area corresponding to P2 only)

As far as I can see, GradientMergeMosaic used a 100% transparency mask for the top panel and 50% transparency for bottom. Data was captured with an OSC camera. I have tried to merge with image still linear but the result is the same.

Does anyone have an ideia of what is going on that makes that loss of color? I would like to understand the reason behind but I'll also settle for a hint on how to repair.

Thanks and apologies for the long post.

Cheers,
André

 

[jmurphy]

I would recommend using SCRIPT > Mosaic > PhotometricMosaic instead of GradientMergeMosaic (well I would, I am the author!)
It needs linear data to work on. This is critical, so the less processing you do on each panel the better. Ideally use the raw stacked data, and align each panel by using MosaicByCoordinates. The script will then match both the background gradient and the brightness scale.
It is a bit more complicated to use than GradientMergeMosaic, and I would recommend you read the Prerequisites and Quick Start Guide in the help file. However, it should be capable of producing excellent results.
Regards, John Murphy

 

[fredvanner]

WARNING: novice talking beyond his experience...
... I would be very cautious about applying DBE to separate, overlapping images, particularly in this region where there is hardly any "real" background, if you expect the resulting bacgrounds to match. The background generated by DBE is critically dependent on the sample points, so it can't do the same background subtraction for both images.

 

[niteman1946]

Hi Fred,
your caution is interesting. My practice on mosaics is to take the panels through integration, cropping and DBE or ABE. And then begin the merging.
I don't recall having problems with background previously, but I've slept since then. I'll keep your comments in mind for my next mosaic.

Thanks,
Mark

 

[fredvanner]

This region is so full of delicate backgrounds with subtle colours that I would expect any background extraction to be risky. I think ABE is a non-starter. DBE will be dependent on picking samples that really are background, not faint nebulosity of one colour or another. One check would be to closely examine the DBE modelled background, and reject it if it has any colour in it that looks like the real (faint nebulosity) colour in that region. Keep rejecting / reselecting DBE samples until it is definitely not subtracting out the colour you are trying to image!

 

[Andre Vilhena]

This region is so full of delicate backgrounds with subtle colours that I would expect any background extraction to be risky. I think ABE is a non-starter. DBE will be dependent on picking samples that really are background, not faint nebulosity of one colour or another. One check would be to closely examine the DBE modelled background, and reject it if it has any colour in it that looks like the real (faint nebulosity) colour in that region. Keep rejecting / reselecting DBE samples until it is definitely not subtracting out the colour you are trying to image!

Hi Fred,

Thanks for your notes and I truly agree with you. Actually, the first DBE done had one sample ill placed and indeed removed a lot of the red color. When I noticed it, I re-did the DBE and result was much better.
But in this specific case, I don't think it's the root cause because I can get the nice red colors in the subframe just before merging it (see left photo above) and I only loose it after the merge (see right photo afterwards). This means that after DBE the colours are there and loss is done during the merge process.

Cheers,
Andre