Registering Individual LRGB Image Workflow?

[twade]

To all,

I'm working on a set of individual L, R, G, and B images of the Rho-Ophiuchus region.  What's the recommended procedure to register all the images?  For example, is it best to register them on a color channel basis (e.g. all Red images get registered together) and then register them in final form (LRGB)?  Or is it best to register all images at the same time using one reference image?  All my L images are binned 1x1.  All my R, G, B, images are binned 2x2.  The low elevation (19 degrees) and limited time forced me to go LRGB. 

I apologize if this has been answered before.  I did a search but didn't really find anything.

Thanks,

Wade

[Philip de Louraille]

The least amount of registration, the less "noise/distortion/artifacts" added by the program. My opinion? Just do one. The fact that you have some binned ones is not an obstacle. PI is smart enough to figure out what's going on.

[Carlos Milovic]

I would go using one of the L images as reference...

[Niall Saunders]

Hi,

The response that I had from Juan when I asked/mentioned this some months ago was along the lines that the 'distortion artifacts' introduced by multiple stages of registration (i.e. align all individual channels to 'their reference', then realign all the individual masters to 'their reference') will NOT be that significant.

And he is probably right.

However, the 'purist in me' is willing to create an image called 'MasterAlign' and to use this to align every other image once, and once only. Now, MasterAlign could just be a copy of a selected image, ot it could even be an image from the likes of the StarGenerator process.

What this gives me is an 'understandable image' when I come to re-use, or re-investigate, a ProcessIcon set some time in the future.

Cheers,

[twade]

To all,

Thanks for the advice.  I went ahead and kept them together; however, next time, I'll experiment with doing them all at once.  I have placed a preliminary image in the Gallery.  I still have some work to do on it but I am quite pleased with the results, especially considering how low it is at 45N.

Wade

[Andres.Pozo]

I usually create a reference image registering and integrating the L frames. Then I register all the images (including the original L frames again) using the reference frame.

[Jack Harvey]

Like Andres I select one image with a low fwhm and good aspect ration, copy the image and rename it REGISTRATION.  All subs from all filters are registered to this image.  Of course this REGISTRATION is not included in integration.

[twade]

Jack,

Of course this REGISTRATION is not included in integration.

Why don't you include it?  I realize it will be "different" than all the other frames, but it still should help with the S/N, especially with fewer images.  I generally don't have the clear nights to take more than 6-12 images so sacrificing one image is painful.  Is there any research out there that shows you "gain" by not including your REGISTRATION image?

Thanks,

Wade

[Niall Saunders]

Wade,

I think what Jack means is that, say you have 10 images (image01, image02, . . ., image10). You establish that, say, image07, is the 'best' image of the bunch. You COPY that image, and rename it as RegistrationMaster (or whatever). This STILL leaves you with the first 10 images - which you then align against the RegMaster. Obviously, image07, being identical, will not 'suffer' during the registration process - but it WILL be included.

So, you still end up with ten 'aligned' images, ready for further processing - such as ImageIntegration.

You also now have a specifically named 'master' image that you can add to the other 'master' images that you generate as part of your pre-calibration steps (MasterBias, MasterDark, etc.)

Is that any clearer?

The only downside - if it is one - is that you sacrifice a tiny bit of HDD space to hold an 'identical copy' of only ONE of your other images. But it does have the advantage that it then becomes absolutely clear what the 'position reference' for the final image is - and it can even help get your scope alignment 'just right' if you have to split imaging over multiple sessions.

Cheers,

[Jack Harvey]

Niall has it correct - as usual<G>.  THe REGISTRATION is a copy of an image in the stack to be aligned and integrated.  Does not help to integrate a copy with the original.

[sleshin]

Using Naill's example, why make a copy of image 7 rather than just using the original image 7 as the reference image for registration. That is, is there an advantage to using a copy of the image rather than the original?

Steve

[twade]

Jack and Niall,

Oh, I see now.  It's a good way of knowing exactly which image you used for registering all other images.  Since it is just a copy, the original still gets integrated into the final image.

Thanks for the clarification,

Wade

[Jack Harvey]

You got it Pardner!

[Niall Saunders]

s there an advantage to using a copy of the image rather than the original?

No - the ONLY advantage is 'cosmetic' - when you come back to revisit the data in three month's time (or three hours, in my case ), then you have a better chance of identifying which image it was that you uses as a "RegistrationMaster"

And, as I said, the ONLY penalty is a few bytes of cheap-as-chips storage area.

Jack's idea is certainly one that I am going to adopt from now on - it also makes re-using saved ProcessIcons more 'intuitive'.

(What do you mean "you never bother saving your process steps"? ONLY a 'True Jedi' can memorise every single step that was done to get from RAW to APOD. Time to set the Pixel Police on you )

Cheers,

[Jack Harvey]

Sometimes i find I might have all the Ha data and not all the RGB data.  In doing the preliminary production of a Ha master I make the RegistrationMaster and then in a week or two (maybe a month when I finish collecting the RGB data I know exactly which frame to use to register the RGB frames too.

Certainly not a requirement, but it is a standard convention that works well for me.