LRGB processing causing weird stars

Ronald Clanton

Adam,

I've been transitioning to taking LRBG mono images rather than OSC images.  I'm afraid that the process is showing some cracks in my understanding/learning.  I've watched all your videos on LRGB, but can't seem to find I'm looking for... probably is in there somewhere, but there is a lot of material!

Anyway, I'm processing LRGB mono images using BIN 1x1 for lum and BIN 2x2 for RGB (my first attempt at using mixed resolutions).  When I process through WBPP, I am getting weird stars... ringing, etc.  I've walked through the process files to identify where the problem starts. Calibration and Cosmetic Correction look good, but in the Registration process it changes my BIN 2x2 images to BIN 1x1 resolution and this seems to be where the problem begins.

So... a couple of questions:

1. I always thought that the BIN 2x2 images would retain their resolution through the LRGB Combination step and then bring in the lum image to increase details at the BIN 2x2 resolution. Clearly I'm wrong, but I would like to understand why?

2. Since this is causing me problems with my stars, do I need to resample my lum frames to BIN 2x2 before WBPP? or is there a way to handle all of this within WBPP?

3. Is using mixed resolutions a bad idea?

I can upload some frames to DropBox if that will help.

Thanks,

Ron

BlockHead

HI Ron,

I simply don't have time to reprocess everyone's data !! LOL

I think you need to show me some images of the ringing you are talking about.

Concerning your questions... 

1. No. You are supposed to have the 2x2 upsampled to match the Lum. This is correct.When you use the LRGB blending (LRGBCOmbination).. you are colorizing your Lum with your RGB data. Both the L and RGB data are stretched. If you overstretch.. you can run into problems. Again... pictures will be helpful.

2. No. Allowing WBPP to register your 2x2 to your 1x1 data is perfect and correct. It is handled correctly in WBPP.

3. No... it isn't bad. However, it does some with some stuff you have to deal with (that you are now beginning to understand). It used to be that binning saved time in acquisition because you minimize the readnoise when binning. This isn't really true anymore with CMOS sensors. Binning really offers no benefit.

It is better to acquire 1x1 and create a luminance from the combination of all of the data (all the masters of RGB). 

-the Blockhead

Ronald Clanton

Adam,

LOL!  No, I don't want you to process my images.  I've attached a screen capture of the stars and a link to the master files of the LRGB.

Going forward, I think I will stick to 1x1 only.  I didn't have any problems with a prior image when the resolution all matched.  This problem is more about me trying to learn more of how things work and to maximize what the OTA, camera and PI can do.

Bottom line.... WBPP worked as expected... Likely, I need to fix some Registration setting.

Link to files...

https://www.dropbox.com/scl/fo/lnz7to58upv58ic16g895/h?rlkey=4mgujwt4t8aj604rdjcp8eieb&dl=0

Thanks,

Ron

BlockHead

The artifact you are seeing likely comes from the Lanczos interpolation method of StarAlignment. You might cnosider using bicubic b-spline if this disturbs you.

-the Blockhead

Ronald Clanton

Wow!  What a difference in the stars... total fix.

Is there a down side to using bicubic b-spline?  It seems so superior to Lanczos.

Thanks!

Ron

BlockHead

It will slightly enlarge stars... it is a function that smears things a bit more by the interpolation. However, for undersampled data, this is rarely a problem.

And... look at that... just a single picture and I was able to offer a reasonable suggestion. :)

-the Blockhead