Star Color Calibration ... very Magenta 🙂

Just a screenshot to show the results: Same on my Mac as on my Windows 7 PC

Different boxes do not change the result. Tried calbrate background again, same. Same issue with my Crab Nebula image from yesterday. Might be the workflow -> Combine LRGB images combined with APP -> LP Removal -> CBG -> CSC -> Magenta. Never had that problem with my Canon.

P.S. Yes there is a lot of coma, still trying to figure out the best working distance of my GPU to my ASI  🙂 

Thanks Mabula 🙂

Hi Frank,

Thank you for sharing this result.

I have downloaded your file.

I agree, the result is too purple no matter which stars you select. The multipliers reported in the Star Color Calibration module seem to overshoot to magenta.

R = 1.055

G = 0,88

B = 0,97

The RGB combine tool can be used to compensate, so I tried with the settings shown in the screenshot, just directly on your file:

So I use

R = 1,025

G= 0,96

B = 1.025

And mix some of the blue and red channels into the green channel.

I do think I know why this happens so I need to improve here, work on the module is not finished yet and I will use your image as well for further testing 😉

Kind regards,

Mabula

Thanks Mabula, quite tricky to get this right. I imaging this is quite an algorithm 🙂 Keep up the great work!

Thank you Frank, yes, it's not simple to get it right for a large range of images. Currently the calibration is done on only 3 parameters. That keeps it simple, but also very limited. I am working to calibrate the colors with 9 parameters (and possible even corrections for both atmospheric and galactic extinction).

So to be continued..., I'll use your image for testing as well, so thank you 😉

Mabula

Great, looking forward to it. In case you need the stacked and registered LRGB data I used for the combine, please find the dropbox link below:

Dropbox File (Private or Invalid)

In the combine tool I used the defaults of LRGB 100% each.

Thanks Frank, I downloaded the data 😉

Mabula

Hi Mabula,

I too seem to be getting strong magenta colouring in one my current work-in-progress image of IC405, see attached, using v1.055.  I 've tried to correct using HSL without managing to get a significantly improved result.   I've tried going back to the RGB Combine stage but also ended up with a rather similar result.  

Do you have any suggestions as to how I might achieve an improved result with or should I wait for 1.056 which I take will be released shortly? 

Happy to provide LRGBHa stacks if this would be of use.

Best regards

Mike

Hi Mike,

This seems completely magenta too me ;-(

Do provide the LRGBHa stacks and I will have a look on how to combine these in the RGB combine tool and subsequent color calibration.

1.056 is soon to be released with adjusted star color calibration, but I think this issue might have to do as well with how you combine the individual stacks to an initial composite.

You can send the stacks to me using wetransfer and to e-mail address: support@astropixelprocessor.com

Kind regards,

Mabula

Hi Mabula,

Just to follow-up my previous post, I took the above .jpg file into Photoshop and after some heavy adjustments in Magenta-Green and lighter adjustments in Blue and Reds, I obtained this new result.  

With some careful reprocessing I am hoping I might be able to recover a little more reflected Blue light in the nebula highlights and in PS Levels I noticed a small loss of highlight detail: but apart from these small points I have to say I very much very much pleased by this result which still owes a great deal to the cap

abilities and ease-of-use of APP.

Regards

Mike

Thank you Mike,

It should help you to preserve more details in the highlights if you save your integration with a milder stretch to a TIFF file before processing in PS I think.

I am working on the star color calibration to improve this. On some images it wokred very nice, but on some the results became way too magenta, like yours. So I am working to improve this.

Thnak you for your nice compliments 😉

Mabula

Hi Mabula,

As requested, hese are my HaLRGB stacks for IC405.

I think my problem probably lies with the Blue stack which is much denser than the others.  During capture I noticed that the average pixel value in the Blue subs much lower than in all the other subs.  Maybe I should increase my exposure times for my Blue frames to compensate for lower CCD sensitivity in Blue?

During RGB_Combine my approach has been to use the sliders to get the histogram peaks in RG&B approximately aligned (which in this instance has required very considerable adjustment) and then to add L and Ha to bring up the image details.   Is this the right approach or am I taking a wrong direction right at the start?

I appreciate your interest and assistance.

Mike

Hopefully this Dropbox link works, I do not use the facility very often:

Dropbox File (Private or Invalid)

Hi Mike, thank, I am downloading now, will report back later today 😉

Mabula

Hi Mike, 

Apologies for the delay, I'll try to give you feedback first thing tomorrow 😉

I am rather occupied with the upcoming release 1.056. I am happy to report that the star color calibration routine will get a very big upgrade after weeks of hard work. The new module will contian 3 models

• the old model using only 3 parameters
• a new model that models the star colors to match black body objects (stars radiate on broadband wavelengths as black bodies 😉 ), so this model uses  real physics. 9 parameters
• And this model with an extension to model atmospheric extinction as well, 12 parameters, this will correct both star colors and the sky background for atmospheric extinction.

To further improve control, for each model, you will be able to fine-tune red/blue ratios of the stars and you can fine-tune any green/magenta cast to get rid of it 😉 along with 2 graphs that will show the star colors of the selected populations giving you a feel of how the model is performing.

I am really happy with the test results, so I will try to release it tomorrow...

Mabula

Wow really looking forward to it!

Didn't complete the final changes yet, but nearly there.. 😉

APP 1.056 should be released tomorrow (or I'll start with eating my left shoe... 😉 )...

And I'll report back on your data Mike !

Mabula

You sure know what you're doing with these algorithms. I am really looking forward to the update also. I've been reprocessing some old images  with some outstanding results  I can't wait till I get better at Imaging and I'm able to use this software at its fullest it's really good software.  And I have a good idea for your new star color algorithms, how about writing an algorithm for each existing imaging camera? That shouldn't take too long.

Hi Greg,

And I have a good idea for your new star color algorithms, how about writing an algorithm for each existing imaging camera? That shouldn't take too long.

Actually, the new star color calibration module uses real physics that will adept to the filters that were used in acquiring the data.

So the calibration will work with different filters and camera's. It will adept to the filters width and central frequency/wavelength. You will need to model the star population on a line 😉 by adjusting the slope and constant of a function y = a*x + b, so you can tweak a and b in such a way that the stars will lie as much on such a line. This modelling is the correction for the filters width and central frequency following how light of a black body light source is recorded by your camera using any filters in the optical path... 😉

A teaser:

Mabula

Posted by: Mabula Haverkamp - Admin

Didn't complete the final changes yet, but nearly there.. 😉

APP 1.056 should be released tomorrow (or I'll start with eating my left shoe... 😉 )...

And I'll report back on your data Mike !

Mabula

So how is the taste of your left shoe ?  😉 😉 😉

Hi Mabula,

Your advanced publicity for v1.056 has aroused considerable interest from me and it seems many others.  I'm retired now but I well remember the pressure that a target date for a new release creates.  However, from a reputational point of view I think it is most important for you to get the new release functioning reliably and in the way that you have envisaged.   Astronomical objects have been around for a considerable time so I am more than willing to wait a few more days in eagar anticipation than to download a version that  has not been well tested. 

I hope you get the outstanding issues with your new release resolved and tested to your high standards in the next few days. 

My images can easily wait a short while longer.

Strength to your elbow

Mike

Posted by: mestutters

Hi Mabula,

As requested, hese are my HaLRGB stacks for IC405.

I think my problem probably lies with the Blue stack which is much denser than the others.  During capture I noticed that the average pixel value in the Blue subs much lower than in all the other subs.  Maybe I should increase my exposure times for my Blue frames to compensate for lower CCD sensitivity in Blue?

During RGB_Combine my approach has been to use the sliders to get the histogram peaks in RG&B approximately aligned (which in this instance has required very considerable adjustment) and then to add L and Ha to bring up the image details.   Is this the right approach or am I taking a wrong direction right at the start?

I appreciate your interest and assistance.

Mike

Hopefully this Dropbox link works, I do not use the facility very often:

Dropbox File (Private or Invalid)

 

 

Hi Mike,

Finally, working on your data, first I processed the blue lights with matching calibration frames:

I noticed that the masterdark still had the bias pedestal in it, because the median value of the masterdark was a bit higher than the median value of the masterbias.

So I calibrated the lights only with the masterdark, masterflat and the BPM:

This looks good, but I see clear banding...  so I redid a new integration now with calibration of the lights using only the masterbias, masterflat, and BPM:

This looks better, a question pops-up, did you subtract the masterbias from your flats?

I'll now start with your LHARGB combination...

Mabula

Hi Mike,

I am working on your LHARGB composite, I think it's best if we first focus on getting good calibration since I need to apply big adjustments with the "remove light pollution" due to bad flat-field calibration.

The integrations don't seem to be well flat-field corrected. If you followed the same path for the R, G, L, Ha integration as the B integration, then I think we can improve a lot here ;-).

As shown in my previous post, the bad blue stack that you got was the result of subtracting both the masterbias and the masterdark from the lights while the masterdark wasn't bias subtracted. This will then subtract the bias pedestal/offset twice.

So I think it's best to focus on getting good calibration first 😉 before any other processing.

Kind regards,

Mabula

Hi Mabula,

When I prepared my calibration subs I believed I had correctly followed the steps described in your Data Calibration Principles post, however it would seem that at some point I must have made a small mistake.  Loading my MB and MD frames into SGP I observe the following stats:

                               Median  Min     Max

MBias                       620      537      14432

MD 150s                     4           1        45817

MD 300s                  639     625       50722

MD 600s                    6           1        50405

MD 1800s                  7            1        56376

So it would seem I made a mistake somewhere in preparing the 300s MD.  I will redo this frame and forward to you shortly.  

Next time I redo my MDs I will know how to perform a check.

Mike

Hi,

I've prepared a new 300s MD (attached) and observe the following pixel stats in SGP, which I assume are more in-line with what you would have anticipated:

                      Median Min    Max

MBias               620    537    14432

MD 300s         639     625   50722  (old)

MD 300s           20       1      45790  (new)

The median pixel value for the new MD looks slightly out of line with the other MDs (details in previous post and prepared with the previous release of APP I believe).  I will check each of the dark subs individually to see if any of these are at all unusual.

I'm interested to see how the stacked lights differ using the new MD but may not get round to doing this until tomorrow.

Mike

Hi Mabula,

I've been working with v1.056 getting a little more familiar with the new CSC controls.  Below is my latest attempt at IC405.  I think the star colours are looking pretty good but the red is still looking rather too magenta and I would like to see a some more bright blue of the reflected light of AE Auriga in the brightest areas of the nebula. 

Suggestions welcome.

Mike

Posted by: Betelgeuze

Posted by: Mabula Haverkamp - Admin

Didn't complete the final changes yet, but nearly there.. 😉

APP 1.056 should be released tomorrow (or I'll start with eating my left shoe... 😉 )...

And I'll report back on your data Mike !

Mabula

So how is the taste of your left shoe ?  😉 😉 😉

It was pretty bad actually ;-( bought new ones today...

Posted by: mestutters

Hi Mabula,

Your advanced publicity for v1.056 has aroused considerable interest from me and it seems many others.  I'm retired now but I well remember the pressure that a target date for a new release creates.  However, from a reputational point of view I think it is most important for you to get the new release functioning reliably and in the way that you have envisaged.   Astronomical objects have been around for a considerable time so I am more than willing to wait a few more days in eagar anticipation than to download a version that  has not been well tested. 

I hope you get the outstanding issues with your new release resolved and tested to your high standards in the next few days. 

My images can easily wait a short while longer.

Strength to your elbow

Mike

Thanks Mike,

You are completely right, I did extensive testing before releasing. It can be rather time consuming, but is really important indeed 😉

Thank you,

Mabula

Posted by: mestutters

Hi Mabula,

When I prepared my calibration subs I believed I had correctly followed the steps described in your Data Calibration Principles post, however it would seem that at some point I must have made a small mistake.  Loading my MB and MD frames into SGP I observe the following stats:

                               Median  Min     Max

MBias                       620      537      14432

MD 150s                     4           1        45817

MD 300s                  639     625       50722

MD 600s                    6           1        50405

MD 1800s                  7            1        56376

So it would seem I made a mistake somewhere in preparing the 300s MD.  I will redo this frame and forward to you shortly.  

Next time I redo my MDs I will know how to perform a check.

Mike

Excellent Mike 😉

Hi Mike,

Were you able to get good calibration for  the Green, Red and L channels now?

These 3 look like something went wrong as well:

The Ha looks much better:

So before fixing the composite, we need to get these light properly calibrated I feel 😉 If the flats are to blame then it is like it is off course.

Mabula

Hi Mabula,

I'm not certain that calibration is the issue here.  I noticed long ago that the RG subs were poor in comparison with B and Ha.  Maybe it is LP or I was unlucky with seein when I cas capturing the poor subs despite them being acquired over several nights and in sessions when I also captured B subs.  Hopefully I will be soon get a chance to capture some more subs in the near future,

Anyway, all is not lost.  I've also been working on M52/Bubble Neb and this to me is looking very good.  This is only the result of first quick pass with v1.056, just to see what I was getting.

Regards

Mike