Page 1 of 1

Narrowband processing and general questions

Posted: Tue Jan 24, 2017 11:35 am
by Notty
Please feel free to merge this with the similar thread on the same subject, but as that one seems to have gone cold I thought I'd start afresh. Background is DSLR, 5y elapsed experience but limited hands on in that time. Previous Photoshop user. I've not moved to mono CMOS cooled cam w.filters and had a chance this week to have a play. Moonlight so went for NB. Attempted a Hubble palatte on the Elephants Trunk and California news respectively. Here are a list of my questions, if anyone could help that would be great. Naiive and not much capture time and I went for same amount of each channel, which I suspect is not enough!

1. I have heard it said that with NB processing it is worth pre-stretching the weaker channels before merging. My data suggests this might be sensible as my Sii and Oiii are way weaker compared to Ha as one would expect. Maybe I wouldn't need to pre-stretch the Ha as it's already a strong signal. In Star Tools would we process each channel separately thus, then re-import in the RGB module, or would we do something clever in layers/masks? (my weakest subject!)

2. Colours. I can't seem to manipulate my colours to get the lovely golds and blues in the Hubble palette via ST's colour module. In Photoshop NB tutorials I've seen, you are able to manipulate the individual colour channels within themselves... does this feed into the question above whereby we would process each channel separately? If so, I imagine it would be extremely hard to gauge the effect of changing one colour on the summed effect when they are merged?

3. Is the "Cap Green" setting only relevant to LRGB imagine as we are playing with visible wavelengths? Obviously most of my image is green due to the dominance of the Ha.

4. General ST question: I can not and have never been able to use the Deconvolution module, as I can't ever get ST to create a mask for me. The alternative seems to me to create a manual one by individually selecting every single star on my image, which can't be right so I must be missing something in my understanding! I have even read the section in the unofficial user manual and am still none the wiser!

5. I have read the very helpful post on the Autodev module and am I correct that it should only be used as a temporary reveal for those of us with less than perfect data? This seems to be born out by my experience and I use Manual Dev, but can someone explain if the GAMMA is used with or instead of the DEV slide? I understand GAMMA is a preset shaped curve stretch applied, but it always seems to wash out my data so I'm not sure what the point of it is?

6. Hue/Saturation? For colours. I assume these are the function in the "LIFE" module, but none of the preset buttons seem to increase the hue or saturation, and invariably make my image look worse!

I tried to attach the best I've managed of the Elephant but once I've attached it and loaded it on the form, if I press submit the BB software just clears the screen of my entire post. Anyway I'm happy enough for a first attempt but it'd be good to know if my colour control is limited by my data or lack of technical skill on Star Tools. Thanks for any help in advance.

Andy

Re: Narrowband processing and general questions

Posted: Tue Jan 24, 2017 6:11 pm
by ChrisLX200
Notty wrote:Please feel free to merge this with the similar thread on the same subject, but as that one seems to have gone cold I thought I'd start afresh. Background is DSLR, 5y elapsed experience but limited hands on in that time. Previous Photoshop user. I've not moved to mono CMOS cooled cam w.filters and had a chance this week to have a play. Moonlight so went for NB. Attempted a Hubble palatte on the Elephants Trunk and California news respectively. Here are a list of my questions, if anyone could help that would be great. Naiive and not much capture time and I went for same amount of each channel, which I suspect is not enough!
Moonlight causes a lot of problems for LRGB but narrow-band is not quite immune either - Ha is the least affected while OIII can be ruined easily by gradients caused by moonlight. Background levels increase too so combining channels taken under moonlight with others with no moon present can result in a background colour bias which you have to compensate for. Generally speaking, I take longer OIII and SII subs than Ha subs - 30min and 20min respectively.

1. I have heard it said that with NB processing it is worth pre-stretching the weaker channels before merging. My data suggests this might be sensible as my Sii and Oiii are way weaker compared to Ha as one would expect. Maybe I wouldn't need to pre-stretch the Ha as it's already a strong signal. In Star Tools would we process each channel separately thus, then re-import in the RGB module, or would we do something clever in layers/masks? (my weakest subject!)

You really need to normalise the background levels first, then you could stretch the signal if one channel is particularly weak. You can also reduce the Ha brightness before combining then adjust the master afterwards. Yes, process each channel separately, combine, then tweak the final result. Remember that colour-mapping is not just a simple substitution of Ha=Green, SII=Red and OIII=Blue - the colours are (or should be) blended such that each primary colour consists of percentages of each NB channel. There is no fixed rule for what these percentages are though and each image is processed individually using unique combinations - it all depends on the strength of the signals.

Caveat: I do use the LRGB function for creating bi-colour NB images (Ha + OIII) because it seems to work very well for that. However, it's usually necessary to boost the OIII signal level - sometimes dramatically - for this to work. Note that simply increasing the saturation level will NOT work to boost (for e.g.,) the blue component, you have to adjust the signal levels with something like the PhotoShop Curves tool prior to channel combination. The LRGB module will create a synthetic Green channel when run with just those two channels (Ha=Red and OIII=Blue) and the result is a nice golden/yellow and blue. This is not Hubble Palette though.


2. Colours. I can't seem to manipulate my colours to get the lovely golds and blues in the Hubble palette via ST's colour module. In Photoshop NB tutorials I've seen, you are able to manipulate the individual colour channels within themselves... does this feed into the question above whereby we would process each channel separately? If so, I imagine it would be extremely hard to gauge the effect of changing one colour on the summed effect when they are merged?
It's tricky, and I don't use ST for this job - for LRGB yes, but not NB. I use PixInsight (pixel math) for tri-channel combination.

3. Is the "Cap Green" setting only relevant to LRGB imagine as we are playing with visible wavelengths? Obviously most of my image is green due to the dominance of the Ha.
As you note - Ha when mapped to Green will result in that colour dominating the image. My normal procedure is to kill the Green using SCNR in PixInsight, (Hasta LaVista Green in PhotoShop does a reasonable job too). Not sure whether ST CapGreen will substitute but seeing as the channel combination procedure is being carried out elsewhere in a different program I have not experimented. After killing the Green it's necessary to adjust the colour balance/saturation again.


4. General ST question: I can not and have never been able to use the Deconvolution module, as I can't ever get ST to create a mask for me. The alternative seems to me to create a manual one by individually selecting every single star on my image, which can't be right so I must be missing something in my understanding! I have even read the section in the unofficial user manual and am still none the wiser!
Creating a star mask is one of StarTools' strong points! You need to practice using the sliders to set the selection process. Mask/Auto/Stars, then I would normally set the 'Feature Size' slider to maximum, and adjust the 'Filter Sensitivity' slider according to the data (values 5 - 12) the lower the number the more sensitive it is to isolating small/faint features, so if the background is noisy you need to incease the value until it's mainly the stars that are being picked up. If the large stars are still being ignored (and there are not many of them) you can just click on them manually to include in the mask, otherwise use the 'Fat Star' algorithm and set the mask behaviour to Add to the existing mask rather than replace. Finally, if you are going to use deconvolution the mask must totally cover the stars - so click 'Grow Mask' to make certain. Save the mask because you will need it again later, then invert it.

For deconvolution you need to set the Radius (typically less than 2), the number of iterations (I usually use up to 20), and the Mask Fuzz (try values from 6 to 8 pixels). Select a small area by dragging if you are just testing the effects - it can take a long time for the full image.


5. I have read the very helpful post on the Autodev module and am I correct that it should only be used as a temporary reveal for those of us with less than perfect data? This seems to be born out by my experience and I use Manual Dev, but can someone explain if the GAMMA is used with or instead of the DEV slide? I understand GAMMA is a preset shaped curve stretch applied, but it always seems to wash out my data so I'm not sure what the point of it is?
Pretty much how I use it, autodev I use initially in case I need to crop stacking artifacts, after that I only use manual develop. I will then use the Wipe module (sometimes with low values down to 10-15 percent just to remove skyglow) followed by a second manual develop (global).


6. Hue/Saturation? For colours. I assume these are the function in the "LIFE" module, but none of the preset buttons seem to increase the hue or saturation, and invariably make my image look worse!
For LRGB colours the Colour module is good, you just have to experiment with the settings as each image will be different. I rarely use the Life module but it can be useful for highlighting the borders of nebulous regions. I don't think StarTools is ever the final tool used to tweak colour balance (neither is PixInsight) - it's usually PhotoShop.


I tried to attach the best I've managed of the Elephant but once I've attached it and loaded it on the form, if I press submit the BB software just clears the screen of my entire post. Anyway I'm happy enough for a first attempt but it'd be good to know if my colour control is limited by my data or lack of technical skill on Star Tools. Thanks for any help in advance.

Andy
ChrisH

Re: Narrowband processing and general questions

Posted: Tue Jan 24, 2017 8:19 pm
by Notty
Chris, thank you so much for taking the time to give me such excellent information. I'm going to digest what you've written and have another play and see what I can achieve. To be fair I'm starting with pretty weak data anyway as I was only giving the new camera 1st light but I was blown away with the Ha I got from a mere 50 mins worth of data low in the sky with pollution and moonlight! I'm slowly getting to grips with Start Tools as well, but my ambition is to one day be able to do everything I need in ST and ditch the (expensive) Photoshop.

Re: Narrowband processing and general questions

Posted: Tue Jan 24, 2017 10:59 pm
by ChrisLX200
Notty wrote:Chris, thank you so much for taking the time to give me such excellent information. I'm going to digest what you've written and have another play and see what I can achieve. To be fair I'm starting with pretty weak data anyway as I was only giving the new camera 1st light but I was blown away with the Ha I got from a mere 50 mins worth of data low in the sky with pollution and moonlight! I'm slowly getting to grips with Start Tools as well, but my ambition is to one day be able to do everything I need in ST and ditch the (expensive) Photoshop.
It would be nice to be able to do everything within one program but to be honest, it's as well to have a number of tools at your disposal - that's no criticism of StarTools which is my No.1 goto image processing program. A very good alternative to PhotoShop is Affinity Photo https://affinity.serif.com/en-gb/photo/ which will do just about anything that PhotoShop can do at a much more affordable price, and it has the same look/feel.

ChrisH

Re: Narrowband processing and general questions

Posted: Thu Jan 26, 2017 12:50 pm
by Notty
ChrisLX200 wrote: You really need to normalise the background levels first, then you could stretch the signal if one channel is particularly weak.
I'm feeling a bit thick here, but I keep reading this section and though I know what it means and what needs to be done, how/what do you use to do this bit of preprocessing Chris? Many thanks again.

Re: Narrowband processing and general questions

Posted: Thu Jan 26, 2017 3:48 pm
by ChrisLX200
Notty wrote:
ChrisLX200 wrote: You really need to normalise the background levels first, then you could stretch the signal if one channel is particularly weak.
I'm feeling a bit thick here, but I keep reading this section and though I know what it means and what needs to be done, how/what do you use to do this bit of preprocessing Chris? Many thanks again.
For normalising backgrounds the easiest is PixInsight - it has a tool just for the job (LinearFit) which will match the background levels of one channel with another, but you can also use the dropper tool in PhotoShop or Affinity Photo and manually adjust the levels so the readout on the background areas for each channel is similar (say, 23 for a mono grey channel, and it would be 23 for each of R,G and B in a OSC - DSLR - image). Visually what you look for is (on a histogram display in PS or AP) for the left-hand slider (which is the black point) to be positioned fractionally to the left of the start of the data - which ensures you do not clip the data by moving it too far to the right. Having done that you can 'stretch' the actual signal - which is lying to the right of the black point you set - by moving the sliders to the left. A better way would be a non-linear stretch done using 'Curves' instead of 'Levels' where you can enhance (or stretch) some parts of the brightenss curve more than others. Always check afterwards that the background has not been altered by the stretch - i.e., still 23. Hope that makes sense - it's not a question of being thick because it's not a simple concept!

ChrisH

Re: Narrowband processing and general questions

Posted: Thu Jan 26, 2017 8:54 pm
by Notty
Thanks Chris, I'm steering well away from Pixinsight for now, but thank you so much for the steer on Affinity Photo. Perhaps that, along with SG Pro, DSS, PHD2 and Star Tools just might be all the complicated software I need... for now!

Re the levels, yes I understood what needed to be done I guess I wondered if there was some fancy way of normalising the levels between them without manually bringing up each image and using "the force" by squeezing the levels sliders in Photoshop, but thinking about it isn't that what the "Colours" module of ST does (with the colours anyway) when you load it, by trying to put match the colour curves for each channel as closely as possible as seen on the graph. I know you're talking about the actual full dataset rather than just the channel set, but if would be cool if something did the same thing for under sampled data (Oiii, Sir) which might need a massive stretch compared to the nice dataset for Ha. I suppose it would also magnify the by definition already high noise with it though, so maybe that's why it's not feasible. Anyway thanks again Chris!