Narrowband Processing First Attempt

[widotje]

Hi all,

Just finished my first attempt to post process narrowband images using DeepSkyStacker and Photoshop.

Ha; 57*300s ; Sii: 21*300s; Oiii:12*300s

45 darks; 20 flats per filter; 30 bias

camera was an asi1600pro and an APO80MM telescope from TS-service (germany).

 

I've used deepskystacker to separate the narrowband images into groups (main = Ha, group 1 = Sii, group 2=Oii), and added the darks/flats/bias frames.

I first created three separate stacks (Ha, Sii, Oiii) and I combined them using photoshop CC.

I've made two 'tutorial' movies with the limited knowledge i have on how i perform this process, see link below.

 

I have a couple of questions, being new at this. First off, should ik change to other software?

I've always used DSS for post stacking my DSLR images, and PS for post-processing...but other tools might be superior?

Second, i now first stack the separate narrowband images into 3 stacks, after which i combine the stacks in photoshop using the RGB channels (HSO). 

Is that the way to go or is there a better way?

See my final picture on astrobin here:

 

https://www.astrobin.com/364400/C/?nc=user

 

Sorry to ask so many questions...:-).

 

 

 

 

 

 

[vlaiv]

There are a lot of software options to consider, some of them are paid, some are free.

If you are used to DSS, you can stick with it, but I use ImageJ to do my calibration and stacking (and hopefully I'll be switching to my own custom software sometime soon).

I would however recommend to switch PS (not sure which version are you using and if it supports 32bit fits) for latest version of Gimp - 2.10, it is free and has a lot of options, and most importantly it supports fits and 32 bit image formats (older versions of gimp support only 8bit).

As for processing, yes stacking each channel is proper way to begin, only thing that you should take care of is to stack all stack to same reference frame - this will save you trouble of aligning later (DSS does this with groups).

When it comes to combining channels and assigning colors to image there are different "standard" approaches - SHO being one of those, but there are also HOO (bicolor), HSO, etc ...

In general case creating narrow band palette is fairly simple (this is one approach, but there can be many different approaches):

You take your narrow band data, it does not have to be 3 different stack, it can be two, or four, it does not matter, and you do following:

Pick coefficients for each of your stack and for each of R, G and B - something like matrix:

        NB1  NB2   NB3
R       0.2    0.4     0.4
G       0.1   0.7     0.2
B      0.6   0.2    0.2

There is thing called Channel mixer that lets you do this with 3 narrow band images/stacks. Matrix for SHO is simply:

1 0 0
0 1 0
0 0 1

You can do this transform prior to stretching your narrow band data, or you can stretch your narrow band data first, and then apply transform to get color channels (results will be different depending on this order of operations).

Btw, nice image, but I do have one suggestion - try to set black point properly. It seems to me, and I might be mistaken, that blue (or OIII in this case) has slightly higher black point than the rest - that is something to be expected, Light pollution will usually be stronger in OIII than in wavelengths above 650nm so background value of OIII data will be higher than those of Ha and SII. If you inspect top left corner of your image, you will find "blue sky" - and I suspect there should not be any OIII signal there - therefore it should be black - you can use it to "equalize" background levels in the image.

 

[widotje]

6 hours ago, vlaiv said:

Pick coefficients for each of your stack and for each of R, G and B - something like matrix:

        NB1  NB2   NB3
R       0.2    0.4     0.4
G       0.1   0.7     0.2
B      0.6   0.2    0.2

There is thing called Channel mixer that lets you do this with 3 narrow band images/stacks. Matrix for SHO is simply:

1 0 0
0 1 0
0 0 1

Thanks for the detailed advise! I'll check out Gimp, didn't know about that one. 

[Pat Warborg]

On ‎07‎/‎09‎/‎2018 at 13:51, vlaiv said:

There are a lot of software options to consider, some of them are paid, some are free.

If you are used to DSS, you can stick with it, but I use ImageJ to do my calibration and stacking (and hopefully I'll be switching to my own custom software sometime soon).

I would however recommend to switch PS (not sure which version are you using and if it supports 32bit fits) for latest version of Gimp - 2.10, it is free and has a lot of options, and most importantly it supports fits and 32 bit image formats (older versions of gimp support only 8bit).

As for processing, yes stacking each channel is proper way to begin, only thing that you should take care of is to stack all stack to same reference frame - this will save you trouble of aligning later (DSS does this with groups).

When it comes to combining channels and assigning colors to image there are different "standard" approaches - SHO being one of those, but there are also HOO (bicolor), HSO, etc ...

In general case creating narrow band palette is fairly simple (this is one approach, but there can be many different approaches):

You take your narrow band data, it does not have to be 3 different stack, it can be two, or four, it does not matter, and you do following:

Pick coefficients for each of your stack and for each of R, G and B - something like matrix:

        NB1  NB2   NB3
R       0.2    0.4     0.4
G       0.1   0.7     0.2
B      0.6   0.2    0.2

There is thing called Channel mixer that lets you do this with 3 narrow band images/stacks. Matrix for SHO is simply:

1 0 0
0 1 0
0 0 1

You can do this transform prior to stretching your narrow band data, or you can stretch your narrow band data first, and then apply transform to get color channels (results will be different depending on this order of operations).

Btw, nice image, but I do have one suggestion - try to set black point properly. It seems to me, and I might be mistaken, that blue (or OIII in this case) has slightly higher black point than the rest - that is something to be expected, Light pollution will usually be stronger in OIII than in wavelengths above 650nm so background value of OIII data will be higher than those of Ha and SII. If you inspect top left corner of your image, you will find "blue sky" - and I suspect there should not be any OIII signal there - therefore it should be black - you can use it to "equalize" background levels in the image.

 

Hi vlaiv,

There are tutorials around for how to process the typical Hubble palette from Ha, OIII and SII, showing those great gold and turquois tints. Those involve lots of "selective color" operations. Any idea how to do this in GIMP? Can that be achieved with simply applying the right coefficient matrix? Do you happen to have a working one at hand you wouldn't mind to share...?

Thanks and cs,

Pat

[vlaiv]

19 minutes ago, Pat Warborg said:

Hi vlaiv,

There are tutorials around for how to process the typical Hubble palette from Ha, OIII and SII, showing those great gold and turquois tints. Those involve lots of "selective color" operations. Any idea how to do this in GIMP? Can that be achieved with simply applying the right coefficient matrix? Do you happen to have a working one at hand you wouldn't mind to share...?

Thanks and cs,

Pat

Best I can do is describe process I last used when searching for good combination in Gimp. There is a thread on it so you can see results:

So process is as follows:

- Open each of your images as B&W

- Stretch them to bring out detail (this is step that you will probably return to as final coloration depends on how much you stretch each channel)

- Do RGB combine using SHO for example (map S to red, H to green, and O to blue)

- Now do Channel Mixer (Colors / Components), add Green to Red 75-100% (maybe subtract a little blue 10-25%), subtract Blue from Green (go crazy here and do around 50%) and add a bit of Green to Blue as well (also 10-20%) to get that sort of golden palette. Can't remember exact ratios, but you can fiddle with it until you get it right (how much subtraction can you do depends on stretch of each channel).

It took me several attempts to get something decent and this is because H signal in this image is rather strong and since we are mapping it to green there is strong green cast, so you want to get back at stretch phase and maybe be gentler on how much you stretch H - there is tendency to bring out all detail in in channel but this as result has much "stronger" stretch of those components that are already strong in signal - like H in above case, so you might want to ease back on those to get good balance between channels.

One last tip that I can suggest - try using Ha as luminance once you do color compose - this is good in cases where Ha is the strongest channel. You can do this by:

- Stretch Ha independently (make a copy and do "full" stretch on it) so that you can bring out the faintest of the details.

- Take above color composite and decompose to Lab model - this will produce L, a and b components.

- Do color compose in Lab model using a and b from above as a and b for this composite, but use fully stretched Ha as Lum instead of above L

[Pat Warborg]

Thanks a lot! I always thought that must work, but just never tried...

Btw. I know exactly what you mean, when referring to the different stretching! I was trying the wizard nebula and no matter how hard I tried, the OIII was just so weak, I never managed to get a decent blue as you typically see in "Hubble" images - until I read somewhere that it just needs A LOT of stretching. That made a big difference.