Mixed SHO/HOO image - need help from the masters

[drmikevt]

I just finished collected the attached data for the Crescent and related nebula.  The image is a straight SHO combine with no processing at all other than color calibration and stretching.  I really like the data and the rich nebula field, and plan to process that properly in SHO, but I also really like the look of an HOO crescent to better bring out the Ha 'skeleton' and O3 'bubble wrap'. 

I would love some advice on how to best create an image where the nebula is SHO and the crescent itself is HOO.  Any and all ideas and advice will be greatly appreciated. 

Thank you!
Mike

[Niall Saunders]

Hi Mike,

I am just sitting here 'daydreaming', without access to PixInsight as such, but I reckon that judicious use of PixelMath might be all that you need (and a suitable 'mask' to identify the Crescent).

Gather your NB data and build your SHO master image 'as normal'. Take that image and apply a PixelMath expression for the R channel that puts the G channel into the R channel (you now have HHO, but you are not done yet). At the same time, use an expression for the G channel that puts (a copy of) the B channel into the G channel (you now have an HOO image as well as your original SHO image).

Build a (GreyScale) mask to identify the Crescent - this is now your third image.

Now you need another PixelMath Process to 'combine' the SHO and HOO images, under the control of the MSK image (I can't remember if PixelMath can simply be applied to a 'masked' image - if it can, then things become much easier). The expression that I am thinking of would be :-

NEW = ( ( (MSK * HOO) + ((1 - MSK) * SHO) ) / 2 )

However, this is purely a 'thought experiment', and I can't verify anything at the moment. Try it and see if you get any where - and let us know how you get on.

[drmikevt]

Well, it sort of works.  I tried it a bunch of different ways:
 - Your way, above
 - Using H, O, O as RGB in PM and inserting
 - using the expression $T[1] for R and $T[2] for G, B (to re-asign the existing SHO data)

They all produce different, but similar results.  Some look better, but none look 'normal'.  When I apply your method I get the result below.  I get the exact same result if I put your equation into all the channels and apply it.  Just FYI - you can apply PM through a mask - same result.  It is curious that the crescent image looks nothing like a straight HOO combine in channel combination.  There must be some interacting between the 2 images instead of a simple inserting of the new data.  I'm not sure where to go!  It would also be great if part of the expression was able to detect and ignore the background and only replace the higher pixel values - I know this is possible, I just don't know how to do it.

Mike

[Niall Saunders]

Hi Mike,

I would like to have a go at this myself - could you attach low resolution versions of the three images in question (SHO, HOO and MSK) to a thread reply.

That way various folks could 'have a go' and reply by posting their result and the PixelMath XPSM Process Icon showing the expression(s) that they used to achieve their result.

[drmikevt]

Niall

I solved it in another way.  I realized I was doing 2 things wrong:
 - Things were much better behaved after I linear fit the 3 NB images before combining them.
 - I was trying to manipulate the data before color calibration.  Everything worked much bette after using the new PCC tool. 
 - After PCC, I applied the expression
R:  $T[1]
G:  $T[2]
B:  $T[2]
to the SHO image through a mask revealing only the Crescent.  I'm really happy with the attached result.  My goal was to bring out more detail in the wrapper nebula, and it does that.  The attached image is so squashed that you can't see it nearly as well as a full size (of course), but you should be able to see the effect.  I need to go back now and process it properly to see what I can tease out of it.  The attached image is just Channel Combine to SHO, PCC, masked Pixelmath, SCNR, HT, and Curves.  Enjoy and thank you for your help!

If you want I can still post the other files, but as you can see the HOO image is not necessary.

Mike

[Niall Saunders]

Hi Mike,

Well - at least you are getting somewhere, and the image doesn't look too 'crazy'.

Once you are ready, it would still be great to see the images 'before' and 'after', along with a nice write-up. Your approach has more 'artistic' merit than 'scientific' merit - although it is a very fine line. After all, you intention is to highlight key areas of the image in a manner that contrasts them with other data present in the image - and this certainly seems to be the case, which makes your approach well worth all the trouble.

I am not a NB imager myself, but it would be interesting to hear comments from those who are.