Image registration with 2 cameras - which one to use as reference image

[macnmotion]

Hi. We often shoot an object with more than one system, and that sometimes means with different cameras. For example on our 4-scope array, we have 2 PL16803 and 2 ML29050.

When it comes time to register the images, I'm wondering which camera to select the reference image from. Based on the following information, can anyone tell me how to make the best selection, resulting in the least loss of resolution?

First, here are the camera specs (the cameras are all on identical telescopes):

PL16803:
16 megapixels (4096x4096)
pixel size: 9 microns

ML29050:
29 megapixels (6644x4452)
pixel size: 5.5 microns

Here is a screenshot from SkyX showing the FOVs of the two cameras. As you can see, a portion of the 16803 will ultimately be lost.



I'd love to understand the calculations that would determine whether a sub from the 16803 or a sub from the 29050 is used as the reference image. Should the 29050 subs be downsampled to match the 16803? Or should the 16803 subs be upsampled to match the 29050? Upsampling may compromise resolution. Up until this point, I've been using a 29050 sub as the reference image, but without knowing if that has been the correct decision.

Thanks for any input. Andy

[RickS]

I often combine data from a couple of different scopes and what I'd probably do in this case is drizzle the 16803 data (image scale goes from 2.06 arcsec/pixel to 1.03) to create integrated masters and do non-drizzle integrations of the 29050 data (image scale 1.26 arcsec/pixel.)  Register the 16803 masters against the 29050 masters and then combine with a noise-weighted integration.

That would be my first experiment at least.  I'd probably try a few different things and compare the results...  Are you doing all the filters with all the cameras?  I'd be inclined to use the 29050 for lum since it has the best resolution and the 16803 for RGB.  That's assuming you're doing LRGB.

Cheers,
Rick.

[kolec]

In the past i tryed  combine data from SBIG 8300 (5,4um pixel) with Fli 16803 data(9um pixel).
(Time of photos was  the same- 1200s. )

But quality data from SBIG is much worse like data from FLI16803. (2,7 times biger pixel, full well capacity , dark...)
Combine thist data - the result was worse like combine data from pure FLI 16803

 Result for me :   It is real to do, but only like training, not for real usable result -( for  fli16803 with sbig 8300)

Try measure SNR, noise  , compare and you will see if..

kolec

[macnmotion]

Hi Kolec,

the question for me isn't whether to combine the data. The data combines nicely and we get great results. My only question is whether to res up or res down when registering the frames.

Thanks. Andy

[kolec]

Hi

Im sorry, or if You have great results , nothing to deal with