No Pleasing Colors with APP Star Color

If in the star color calibration it looks too green, you can adjust the Magenta-Green sliders a bit and move those to the left. If you turn on saturation on the right, do the colors become more to your liking?

@vincent-mod

Hi Vincent,

the Green is not the Problem in APP, the opposite is the case. APP takes pretty good care about the Green. But the Results are to reddish. If i dial up the Saturationi still don´t like it. Looks to red to me. I´ve tried to turn the red/blue slider more towards the blue. But this is in my opinion oposite to what the recommendations say. And still, it looks kind of strange to my Eye. With the Standart RGB Method, it looks more natural to me (always admit, that i have to get rid of the green cast then).

CS Frank

Hi Frank @frasax,

Are you performing the Star Color Calibration on the unprocessed integration/stack?

Please let us have a look at the stack, can you upload it to our server:

https://ariesprodstor.astropixelprocessor.com:7001/

login and password: appuser

create a folder with your name and let me know once uploaded, then I will check what I can get with your stack in terms of color 😉 and share with you my findings

Mabula

Ok, so you can also adjust the colors a bit using the HSL selective color tool. The star color calibration should give you a very good starting point at least, unless the data isn’t purely broadband (mixed with narrowband for instance) or when filters are skewing the spectrum.

@mabula-admin

I´ve uploaded two Files. The Stack, and a Lightpollution Corrected Stack. 

Cheers

Frank

Posted by: @frasax

@mabula-admin

I´ve uploaded two Files. The Stack, and a Lightpollution Corrected Stack. 

Cheers

Frank

Hi Frank,

Thank you for uploading your integration file.

I have done Light Pollution correction and then Star Color Calibration and this is what I get, please have a look at the settings that I use:

2 factors play a role here which you need to take into account.

1. your data was processed with Bayer Drizzle and droplets of 1.5, the integration was done on only 32 frames which is rather little to perform Bayer drizzle with that droplet size. So some of the smaller undersampled stars actually are not well drizzled. I think with droplets of 2, it will be quite a bit better and there will be much less noise as well. I am bringing this forward to indicate that this has effects on colors. Bayer Drizzle is using 2 green pixels and only 1 red and 1 blue per 2x2 cfa pattern. So you expect that the green channel is better reproduced than red and blue. This can be seen on the undersampled stars, some have ugly green/cyan borders because of this. With bigger droplet sizes, you can solve this I think, then the reproduction of red and blue channels should become much better with only 32 frames.
2. With such dense star fields to sample from for star color calibration, there will be a big bias in blue stars being detected over red stars because the blue stars are much more luminous. You need to take this into account in the red-blue slider. In this data, the bias is very big indeed. So yes, theoretically and in real life, there are much ! more red stars than blue. Practically... much more blue stars are now detected than red stars in this widefield field of view in the areas where we are sampling stars.

So to get good colors, I actually had to

1. set the Blue - Red slider to -0.10 , where you were setting it at + 0.15 which, yes, made the data way to reddish.
2. And to compensate for the bayer drizzle with the too small droplets where green is much better reconstructed than red and blue, i had to actually remove magenta cast and thus set the  magenta-green sliders to +0.10, thus injecting green. The undersampled stars now have some green borders which is the expected bias for Bayer Drizzle with too small droplets on undersampled stars. (and we can correct those green star borders cosmetically with HSL selective color)

I have looked up B-V star colors with stellarium and compared them to the brightest stars in the field of view and now this star color calibrated result looks very good when compared to scientific B-V values, I checked the B-V values for stars like

Rigil Kentaurus, Hader, Epsilon Centaurus, the 3 Southern Cross stars:

B-V = 0 would be blue

B-V = 0.6 would be like the sun, white/yellow

B-V = 1.0 would be yellow/orange

B-V = 2.0 would be orange/reddish

I think it looks very good now:

But I have also performed HSL selective color, to kill the green + cyan star borders on the mostly undersampled stars:

1) select green, kill saturation by moving sat to -100 and slightly inject magenta which is opposite of green

2) select cyan, move to blue 0,75 and also kill saturation about -0.75

The effect of the HSL selective correction for the Bayer Drizzle bias can be seen best when the images are next to each other, especially when zoomed in in the undersampled stars 😉 :

I think the result is very nice and regarding star colors it is rather accurate as can be simply checked with those B-V values. I do think that the star color calibration will be better if you would have used a bigger drizzle droplet size.

Please let us know what you think about all of this.

Cheers,

Mabula

Hi Mabual,

thank you for your great support, and yes, this looks much better to me 🙂

I´ve tried Bayer Drizzle for the first time, and as a starting point i choosed droplets of 1.5 to see where i get. I will investigate this in the future. It sounds logically to me, that this can lead to green Halos.

Also, it sounds reasonable that sampling plays a role here. But in the End this is a lot to take into account for propper calibraton. If you anyway compare the Pictures to their B-V values, why not implementing a B-V Star Color Calibration?

I will now try to reproduce your results and see, if i get a better feeling for it. What i see in your Color-Color Diagrams, is far from what i understood about the tool in the Tutorials tha i watched. I Need to think more about that.Thanks again for the great work you are doing here!

CS Frank