California Nebula - Kelling Heath Ha and DSLR

[RobertM]

If it's any consolation, PixInsight also does a excellent job of destroying star colour.  It's probably my lack of experience but I found that all the obvious ways of applying some kind of data stretch also changed the R:G:B ratios in the pixels

That certainly needs investigating have you any examples ?  Of course you could always apply asinh stretching within PI if needed but I take it you were using some of the standard tools.  I'll start a PI thread under technical...

[MarkS]

That certainly needs investigating have you any examples ? 

I'll do better than that - I'll create a 16 bit synthetic example TIF tonight - I need some test data in any case. Besides, it's the only way of working out what on earth various processing functions in Iris, DSS, PixInsight, Maxim and (not forgetting) Photoshop are doing.

The image will contain a row of reddish stars (all with precisely the same R:G:B ratio) with different magnitudes from the dimmest up to completely oversaturated - blurred with a Gaussian to simulate atmospheric effects.  Then a similar row of white stars and a similar row of bluish stars.  The whole thing will be given a synthetic RGB colour balance as if taken by a typical DSLR sensor. It will also have a uniform background sky glow and a hidden faint picture that needs to be revealed.

The goal of the exercise will be to use standard tools in each application to correct the RGB balance, remove the background and then stretch the image to reveal the hidden faint picture whilst maintaining correctly coloured stars.  We then document the required processing sequence for each application.

Iris will win this hands down; in Photoshop I bet it's not even possible; the rest will fall somewhere in the middle.

[JonH]

That certainly needs investigating have you any examples ? 

I'll do better than that - I'll create a 16 bit synthetic example TIF tonight - I need some test data in any case. Besides, it's the only way of working out what on earth various processing functions in Iris, DSS, PixInsight, Maxim and (not forgetting) Photoshop are doing.

The image will contain a row of reddish stars (all with precisely the same R:G:B ratio) with different magnitudes from the dimmest up to completely oversaturated - blurred with a Gaussian to simulate atmospheric effects.  Then a similar row of white stars and a similar row of bluish stars.  The whole thing will be given a synthetic RGB colour balance as if taken by a typical DSLR sensor. It will also have a uniform background sky glow and a hidden faint picture that needs to be revealed.

The goal of the exercise will be to use standard tools in each application to correct the RGB balance, remove the background and then stretch the image to reveal the hidden faint picture whilst maintaining correctly coloured stars.  We then document the required processing sequence for each application.

Iris will win this hands down; in Photoshop I bet it's not even possible; the rest will fall somewhere in the middle.

I really like this idea!

Just thought I'd add my two cents worth, with DSS I think it might also be a good idea to uncheck the align RGB option in the stacking settings.

Although I found having this checked made it a little easier and quicker to process the image I also found I could get better results if I did it myself outside of DSS.
Perhaps this is just me but maybe this is also manipulating the image un an undesired way?...

[Mac]

Sounds like a challenge.

Im up for the photoshop bit.

Mac.

[MarkS]

Im up for the photoshop bit.

I'm very happy for you to prove me wrong regarding my negative comments about PS!

[Mike]

Perhaps an Iris and/or PixInsight workshop at one of the Imaging Sessions or perhaps we should actively set a date for another Imaging Workshop?

[Mac]

well i managed to get this from the original tiff, (not including the HA).
Photoshop only.



Cropped and scaled, needs a lot more work, but this was only about 5 mins of playing.
and i was just seeing what data was there.

Mac.

[JonH]

Well this is what i got out of just the DSLR image, I spent quite a bit of time trying to get this right and worked it in both Iris and Photoshop.
I think the colour balance is still a bit out but the star colours are starting to show threw and think i managed to get the worst of the noise under control.
That banding refused to go away entirely though and can still see bits of it in places!

[Mike]

Well this just goes to show Carole that overall you are capturing good data, it is just a case of processing im such a way that it comes out looking good. capturing the raw data is the easy part, it is the processing that is the 'magic arts'.

[JonH]

Well this just goes to show Carole that overall you are capturing good data, it is just a case of processing im such a way that it comes out looking good. capturing the raw data is the easy part, it is the processing that is the 'magic arts'.

I would say the data was ok, it was a pig to work with though and didn't respond very well to processing. There is defiantly something not right with it, I would normally apply a MUCH harsher asinh stretch in Iris to really bring out the star colours but this just didn't like it at all.
I would REALLY like to have another pop at this but stacked in Iris, what Mark has said kinda makes sense to me so a comparison should revel a huge differance!

[Carole]

Thanks for doing that Mac and Jon, I will now do a newly stacked image without using the RGB controls in DSS and I will also post up the individual subs so you can have a "go" in your favourite stacking software. 

Mark I would also be interested to see what you can get from it with Iris, plus has any-one had a "go" at adding the Ha data yet to the DSLR data?

Carole

[RobertM]

Here's my attempt in PI (DSLR only).  I would be tempted to mask the stars and colour enhance them more separately but for this exercise I've only used whole image processing i.e. not processed any part individually.



Full monte: http://farm9.staticflickr.com/8332/8116898225_c5befc3fbf_o.jpg

Process flow for this was:

Canon banding reduction (default options)
Background neutralisation (default options)
Colour calibration (default options)
Multiscale Median Transform (a little tweeking)
Histogram Transformation (defaults copied from 'Screen transfer function' and all colour channels equally stretched)
Colour saturation (a little tweeking)

No masks or any other special processing.

Robert

[JonH]

Robert, please tell me thats not just the DSLR data! 

[Carole]

That's really good Robert, just goes to show what can be done in expert hands.

Carole

[RobertM]

Thanks Carole but in expert hands it would be much better.  Do you want me to try to add the Ha Data ?  I've never done that in PI before so should be good for a laugh.

Robert