SPCC Narrowband

[Gazz1]

Hi Jarmo,
Thought I would start a new thread for SPCC…the BXT thread is 100+ now.

Have done a lot of testing on SPCC with NB filters and I am getting mixed results.

I have tested on 2 NB targets:

1. IC 2944 Running Chicken CHI-6-CCD
2. M16 Eagle CHI-1-CCD (our standard dataset)

I have tested both with HSO palette
(I tried SHO, but it was too difficult to identify changes without Color calibration and removing magenta)

I used “Integration_RGB” for all testing
I only selected “crop” in AI and no other options.

These are the default settings:

I will refer to this image in several posts.

For NB, it is recommended to select:
White reference: Photon Flux
QE curve: Ideal QE curve

I ran testing on IC 2944 first and got reasonable results with the above 2 changes and leaving the rest as is. However, I got very poor results with M16. This turned out to be because of the “Background Neutralization” (BN) option selected by default. It made an improvement for IC 2944 but not for M16. For M16, most of the image turned Black and I’m not sure why?. I tried running the Background Neutralization process separately and it produced the same problem.
I did get reasonable results with M16 if I ran ABE on Integration_RGB before SPCC, It seems that cleaning up the image before running SPCC helps in this case and I got a reasonable result even with BN selected (the default)

In summary:

• None of this has involved setting NB filters yet, just the default RGB ones.
• White reference was set to Photon Flux for NB images
• QE curve set to Ideal QE Curve for NB images
• Background Neutralization is set by default and caused problems with M16 (I’m not sure why). This a problem with Background Neutralization on this image. It was improved by cleaning up the image a bit with ABE prior to SPCC.

Here is Integration_RGB:

Integration_RGB with SPCC and Background Neutralisation selected

SPCC and BN not selected - is way too Red:

ABE first then SPCC with BN selected:

ABE first then SPCC and BN NOT selected:

I’m not sure what to make of any of this?

I will jump ahead and show you how it looks with NB filters applied:

M16 Intgration_RGB Narrowband Filter applied SPCC BN not selected:

M16 Integration_RGB with ABE then SPCC with NB filters and BN not selected:
I think this is the best result:

[Gazz1]

Comparison of Default RGB filters and NB for M16:

Original Image: Integration_RGB with ABE applied:

RGB Filters: ABE before SPCC RGB Filters no BN:

NARROWBAND Filters: ABE before SPCC NB Filters no BN:

There’s not a great difference, but the Background Neutralization problem is impacting the results.
The IC 2944 results were better where BN worked. I will post these soon.

[Gazz1]

SPCC Narrowband Settings:

I left BN selected because I think M16 is an exception, but I’m not sure.

I got the NB setting from the Astrodon site:

The SPCC settings above are for a HSO image. If it was SHO, then SII would map to Red and we would have “671.6”(SII) instead of “656.3” (Ha).

The Narrowband defaults appear as Astrodon HOO when you open SPCC and select “Narrowband filters mode” i.e. R=656.3, G=500.7 and B=500.7. I’m not sure why?

There is a file “filters.xspd” in my Pixinsight/Library folder on my mac. It has a variety of relevant settings

I think that it stores data that can be added/edited by clicking the “Curve Explorer” and “Filter Management” buttons on the SPCC screen **note that these buttons don’t appear in NB mode:

[Gazz1]

A quick look at IC2944 results:

IC2944 Integration_RGB:

SPCC Default settings: a pretty good result

Same as above but with NB filters applied: I’m not seeing a huge difference here between RGB and NB filters.

Image above without BN applied in SPCC (same impact for the RGB filter equivalent): so BN made a big impact for this image.

Comparison of SPCC All defaults and setting White reference to Photon Flux:

All Defaults:

Photon Flux: it may be slightly better but I’m not sure:

White balance charts for SPCC defaults RGB vs SPC Defaults NB: I’m not sure how to read these yet.

RGB:

Narrowband with all defaults:

Narrowband with Photon Flux:

I’m not sure if any of this is helpful?..I have more questions than answers but I will watch the 2 Adam Block videos and see if that helps clarify a few things.

I looked inside the FITS files and couldn’t see any reference to the wavelength of the filters. I guess that makes it challenging to automate for NB. I think the majority of Telescope live NB filters are the same, so that might help? People can always enter values into fields.

At this stage, the M16 Background Neutralization error is a concern, but it may be something simple that I don’t know about.

Garth

[Gazz1]

Adding the 2 Adam Block videos to this thread for reference:
Part 1: PixInsight SpectroPhotometric Color Calibration: Part 1 (Astrometry) - YouTube
Part 2: PixInsight SpectroPhotometric Color Calibration: Part 2 (The SPCC Process) - YouTube

[jarmor]

Wow, that is a lot of testing! I will take a while to understand all this but a really good set of tests.

Jarmo

[jarmor]

I started testing this and ended up adding narrowband SPCC stuff to AutoIntegrate. I have no conclusions myself so I just added whatever you had commented. I have not tested very much so not sure if all works as they should.

I kind of like the Narrowband auto mode. Should help with different combinations as long as you do not mix filters. Now there are Telescope Live filters and CMOS cameras built in.

Here is the commit message:

SPCC for narrowband images

• Added Narrowband auto mode that automatically sets narrowband mode, white reference and filer parameters. Filter parameters are set only when a single filter is used, not a mix or multiple filters. Filter values use Astrodon Astrodon LRGB 2GEN filters values. Option is checked by default.
• Added option to choose white reference between Average Spiral Galaxy and Photon Flux for SPCC
• Added option to set narrowband mode for SPCC
• Added option to select Background neutralization for narrowband mode SPCC. Option is checked by default.
• Added options to set wavelength and bandwidth for red, green and blue filters.
• Changed AutoSTF and Histogram stretch Auto link for channels to use linked for narrowband if color calibration is used.
• AutoContinue tries to find telescope name and can use predefined telescope for image solving.

[Gazz1]

Hi Jarmo,

That’s incredible.
This will make further testing much easier.
I will do some testing tonight on the same datasets to compare results (they should be the same).

I can then play around with a few settings in AI like Local Normalization, Background Neutralization and ABE to see if this has an impact.

I can’t believe that you have done so much so quickly!!!

Garth

[jarmor]

I must admit it was a bit too much for one evening.

One thing I am a bit unsure is how to do linked/unlinked channels in AutoSTF. Now I changed it so that if color calibration is done then we use linked AutoSTF for narrowband. With HSO I liked the result more although it is very red (H).

In SPCC documentation I also saw good results for normal ColorCalibration with SHO data. Need to look into it, my first results were pretty bad but then I was not using linear data for CC. Currently only color calibration you can do for narrowband in AutoIntegrate is SPCC.

Jarmo

[Gazz1]

The videos that I watched used linked STF for narrowband.

I noticed HSO is very Red also.
I didn’t get much time to test this weekend but I will try and test a bit more tonight (I’m in Sydney, Australia and it’s morning here now).

The Narrowband setting are set to SHO by default as 672 - SII. If, I process HSO, it will switch Red to 656 - is this correct? (assuming it is or my HSO and SHO would look the same).

Garth

[Gazz1]

Hi Jarmo,

I think I’ve fixed this now with SCNR

I’m wondering if something is wrong with the HSO images - they were not what I was expecting?
I have been doing some testing using NGC2244 Rosette Nebula CHI-2-CCD and am surprised by the results.

This is with no SPCC, BXT or NXT - just basic settings.
Tested with 1.58 test 8 and 1.57.1 and get the same result.

Here is the Integration_RGB: - HSO

This looks fine to me.

Integration_RGB_noABE - HSO

Also look good

Integration_RGB_onABE_HT - HSO

This was not what I was expecting.
AutoRGB is the same:

It looks like A HOS image that I have done with Rosette before.

[Gazz1]

Here’s a HSO image of Rosette from Oct 22

Different scope.

And a HSO from CHI-1-CCD that I just ran now.

Maybe there’s something wrong with that dataset?

[Gazz1]

Just ran SCNR on it and it looks a bit more like SHO should:

Before:

After:

Sorry for the inconvenience.
Garth

[Gazz1]

Back to testing with IC 2944 Running Chicken HSO:

Default settings for SPCC, no image solving data entered (automatically extracted from FITS files) and no changes to default SPCC options.

Settings:

Result: good result

Settings:

Result: Looks the same as defaults

[Gazz1]

Just ran this on M16 again - remember I was experiencing a Background Neutralisation problem before when running manually.

It worked fine in AutoIntegrate so it must have been something that I was doing.

M16 HSO - Default SPCC settings:

[jarmor]

I do not remember any changes in basic narrowband processing. But always good to check.

With SPCC it uses now uses linked channels with narrowband. Normal default for narrowband is unlinked. Linked channels create very different results. I am not sure which is the right default, though. I guess I need run a few palettes with different settings see how they look.

Narrowband auto mode is supposed to set correct wavelengths based on palette used. So with HSO the H wavelength is set to red but SHO the H wavelength is set to green.

Jarmo

[jarmor]

Here is my trials with different palettes.

• First row is using defaults. No color calibration, Unlinked channels for AutoSTF.
• Second row is using SPCC with Auto channel link mode. It is a new change where AI uses Linked channels for AutoSTF when there is H assigned into red channel, otherwise AI uses Unliked channels.
• Rest I hope can be figured from text in the image.

It would be interesting to know your opinion of the new SPCC default with mixed linked/unlinked channels.

If the image above is too large here is a link to a bigger version.

Palettes_x3 copy.jpg

[Gazz1]

I Jarmo,

I watched this short video on SPCC by PixInsight. It helps to answer many of our questions SPCC — Working with Narrowband Images - YouTube

Key points are:

• For SPCC with Narrowband: set White reference to Photon Flux; QE Curve to ideal QE curve (to determine the intensity ratio of each channels emission bands - this is explained around 2:00 into the video); and Linked Channels for AutoStretch.

• The second part of the video runs SPCC on M42 Orion in SHO and while the ratio of channels is calibrated correctly, there is no contrast between SII and Ha because the Ha is so dominant.
  It appears to me that the author recommends using ColorCalibration after SPCC on the image to make the intensity of the nebula the same in all 3 channels (explained around 7:00). The author selects the nebula itself as the Region of Interest in ColorCalibration to set the White Reference. I don’t know how you would implement that or even if it’s worth bothering. I tried running SPCC without a Region of Interest (because I can’t see how that could be implemented in AI). I then Ran ConorCalibration on the SPCC result without setting the whitepoint as a Region of Interest (once again, because I can’t see how that’s possible in AI) and the result was Not very good at all. I can’t see the point in getting the correct ratio of NB channels using SPCC and then making them the same using ColorCalibration.
  The resulting image has pink stars also.
  I think we get a better result without using SPCC on SHO images and running SCNR and fix stars,

HSO - RGB Combined

HSO - SPCC and Linked STF

HSO - Color Calibration after SPCC - defaults i.e. No White Reference or Background Region of Interest. Also no Structure detection (as recommended in video)

HSO Same as previous image but with Region of Interests

SHO - RGB Combined

SHO - SPCC and Linked STF

SHO - Color Calibration after SPCC - defaults i.e. No White Reference or Background Region of Interest. Also no Structure detection (as recommended in video)

HSO Same as previous image but with Region of Interests

I’m not sure if this is much help but I’m not really seeing the benefits of SPCC for NarrowBand,
It would be great if you get the chance to watch that video and share your thoughts…I may have misunderstood something.

[jarmor]

Hi Garth,

Thanks for checking this out. I will watch the video and also run some own tests with M42 data.

Jarmo

[Gazz1]

Hi Jarmo,

I think I found a small bug.

When processing a LRGB image of M82, AI ends with an error. I think it involves the change for multiple NB files?

It’s because I have the “Use multiple mappings” selected in the Narrowband processing section as part of my default config.

Garth