Stacking Separate R, G, B channels from OSC

Posted by: @readyjetty

↑

Hi, I want to stack separate R, G, B channels from an OSC.  I assume I first save the calibrated frames with "split channels".  Several questions on doing this:

 

1) Do I need to pick a particular debayer option in the "Load" tab or does the selection not matter because you won't be doing any debayering?

 

2) Is the output data the same resolution (number of total pixels) and if so how are the CFA holes filled for the colors that are now gone?

 

3) If I also want to align channels before I split the data out and selecent "Align channels" does it then debayer before splitting it out.  Does this debayering cause some cross contanimation of the different color signals?  How can I prevent this?

 

4) I assume that after this calibration step, I'll have three sets of color lights, one red, green, and blue and I can then just reload each color plane of lights (one at a time) and stack them and create a mono output that only represents the data in that particular color plane with no cross contamination.  Correct?

 

Anything else I miss or should consider when splitting and stacking each color channel like this?

 

Thanks!

Hi Steven,

If you want to stack R,G,B separately from an OSC camera, then yes, calibrated the OSC data and use the split channels option in 2) Calibrate to get the calirbated R,G,B channels.

1) yes, demosaicing is always done in case of splitting the channels otherwise you would have holes in the R,G,B images.

2) yes, same resolution, cfa holes are fillesd by demosaicing the CFA data before splitting the demosaiced data.

3) Yes, it demosaics first, even before aligning the channels, if we woudl not do this, we could not even align the channels because the star intensity profiles per channel would be incomplete 😉 which would give big errors on the star centroids trying to align that. What do you mean by cross contamination? Most advanced demosaic algortihms use the green channel to reconstruct red and blue as well. Simple demosaic algorithms wouldn't but they results would be less sharp as a consequence. The Adaptive Airy Disc algorithm used green to reconstruct red and blue. Bilinear wouldn't but the result is clearly inferior. If you are concerned about the effect of color influence from demosaicing the cfa pixels, then I would suggest the following:

Do not split the channels in 2) Calibrate, make an RGB integration of the OSC data using Bayer/X-Trans Drizzle, then no demosaicing is done and only the CFA are drizzled into the integration giving you the cleanest and most color pure result without the demosaicing effect. Downside of this path is that Drizzle requires dithering of the data and plenty of data because drizzle is a known noise injector due to how it works. With Bayer/X-Trans drizzle, set the drizzle droplet to 2-3 pixels and keep the integration scale at 1.0. To get similar sharpness as demosaicing. If you lower the droplets, it will become sharper, but noisier. (Again, this is how drizzle works 😉 )

4) Yes, correct in the sesne that you will have R,G,B lights ready to make R,G,B mono stacks. No not correct in the sense that the data is demosaiced thus there shall be influence of the color channels on each other allthough it will be minor. Again, if you are concerned about that, you must use the Bayer/X-Trans Drizzle way for sure  😉

In addition, splitting OSC data is only creating extra work here and not adding anything.  You would only want to split the channels if you would want to register and normalize and stack the R,G,B channels alongside mono channels like Ha, O3, SII ;-). Or if you need to perfrom align channels as well due to having clear chromatic aberration in your setup.

Mabula