Star Mask, Star halos & noise

[mcbbcn]

Hi,

I've been exploring Startools, and during my recent LRGB processing, I ended up with 3 issues which I didn't know how to solve & I'll be providing image examples at the end.

1) I think the Autodev drives too much noise during processing, and I was wondering how you tame down the processing. I followed Ivo's recent processing steps shared at CN, but I was still ending up with too much noise.

2) Star Halos. I read Ivo's star halo reduction post, but the halo reduction was affecting just on the smaller stars, even when I had a mask that it was covering the bigge star's halos, it was still not reducing the halos of the bigger stars

3) Star masks look square, and in some instances, it ends up affecting the shape of the halo.

StarTools Image (applying Ivo's CloudyNights HaRGB process steps)

https://picasaweb.google.com/lh/photo/9 ... directlink

StarTools+Pixinsight In Pixinsight I applied Noise reduction, curves transformation (increase red -vs pink- and reduce lightness), and applied starmask and histogram to reduce star halos which I could not completely reduce -see example of 3 blue stars halos at the bottom left side-...

https://picasaweb.google.com/lh/photo/U ... directlink

StarTools+Pixinsight+Photoshop AstroTools I applied auto-contrast, astrotools noise reduction, local DSO contrast enhacement...

https://picasaweb.google.com/lh/photo/0 ... directlink

Ideally, I would like to know how to get better at managing those star halos and reducing the overall noise...

Thanks for your help! Clear Skies!

M.

[admin]

Hi,

I've been exploring Startools, and during my recent LRGB processing, I ended up with 3 issues which I didn't know how to solve & I'll be providing image examples at the end.

1) I think the Autodev drives too much noise during processing, and I was wondering how you tame down the processing. I followed Ivo's recent processing steps shared at CN, but I was still ending up with too much noise.

The AutoDev module can be instructed to ignore finer detail (such as noise) using the 'Ignore Detail <' parameter. Ofcourse, this will not stop AutoDev from allocating dynamic range to noise that sits in the same dynamic range as larger detail.
While Noise may show up visually, Stretching in StarTools (as long as Tracking is on), should minimally impact noise propagation into other modules. The final noise reduction step should be able to get rid of your (smaller scale) noise very effectively, especially since it is scale based. Indeed, looking at the 1st link you posted at a larger scale (e.g. zoomed out), I much prefer it as it shows a whole lot more larger scale detail.
The question that AutoDev solves is; how can I allocate global dynamic range THE most effective to show the maximum amount of detail (that is larger than 'Ignore Detail <') in my image?
It is then up to you to take care of remants of the detail that you *didn't* want to see (e.g. the detail is smaller than 'Ignore Detail <') that shared the same dynamic range with the larger detail that was brought out.

What AutoDev is really showing you here, is that you are suffering from 'resolution greed'. In this instance, your signal is too low to support bringing out the fine detail, while you are willing to throw away (i.e. stretch less) the wonderful bigger detail that you did capture, just for the sake of holding on to your resolution. What you end up with is a rather underwhelming over-smoothened image that is high in resolution but very low in visible detail for the time you put into acquiring your data.
The Bin module allows you to convert 'useless' resolution into more signal and hence, recover and display the larger detail that you did capture. Importantly, the Bin module allows you to adjust this balance freely until you are satisfied with the resolution:signal ratio. I find that once people get their head around the relationship between resolution and signal, they discover that they can go much deeper than they thought they could. The results they then achieve, even with modest equipment, are absolutely fantastic.

Note that,by default, AutoDev does its thing for your whole image, however if there is a particular object of interest in your image, then you can also specify a region of interest.

All that said, if you absolutely must hold on to your resolution you can use the Gamma parameter in the AutoDev module, or use the classic Develop module.

2) Star Halos. I read Ivo's star halo reduction post, but the halo reduction was affecting just on the smaller stars, even when I had a mask that it was covering the bigge star's halos, it was still not reducing the halos of the bigger stars

Any chance you could post a crop of some stars that you'd like to see fixed up? It'd be very helpful!
Also, what further settings did you use?

3) Star masks look square, and in some instances, it ends up affecting the shape of the halo.

The squareness sounds like it may be the result of the 'grow' and/or 'shrink' operations, a star mask itself when it is first generated should not come out square!

Would love to have a play with your data...

[mcbbcn]

Hi Ivo,

Thanks for the quick reply. You're awesome.

As you suggested, here is my data...

I created the Luminance following your CN steps:

http://dl.dropbox.com/u/8971416/Challen ... R_ST2.tiff

Then, I created this RGB following your CN steps:

http://dl.dropbox.com/u/8971416/Challen ... site3.tiff

...and this is an example of the star halos:

http://dl.dropbox.com/u/8971416/Challen ... _halos.tif

If you want the previous data (before RGB combination), I can upload it and share it too...

Also, I'm curious about the sequence to apply the BIN module, do you have any preference for applying before or after any other sequence steps...

By the way, I was processing RGB, I did not set the image to track - it did not say that on your instructions - but it sounds like I should do it when processing Luminance or processing RGB...makes sense...

Look forward to hear more about your processing techniques...Thanks for everything! Clear skies to you!

M.

[admin]

Thanks Miquel, that really helps narrowing down the problem, which is probably this;

By the way, I was processing RGB, I did not set the image to track - it did not say that on your instructions - but it sounds like I should do it when processing Luminance or processing RGB...makes sense...

Whenever you are processing anything that is linear and unstretched, let StarTools know when it asks you. This way it can engage tracking. The only reason I 'lied' to StarTools about the data being non-linear at the start of the tutorial, was because I needed access to the Layer module to synthesise a master luminance frame for your particular data set (so I could blend in the Ha). I did however click the Track button afterwards to engage tracking manually.
By not using Tracking you are missing out on the best bits that StarTools has to offer - notably the extremely effective noise reduction. The latter is (optionally) performed when you end tracking.

Your luminance is really excessively noisy and is also the cause of the 'Tighten' algorithm not performing as well as it normally would. Did you have a separate luminance and/or Ha frame?
What were your exposure times for L, Ha, R, G and B?
I'd be interested to see the source data if that's not a problem...

As for binning with StarTools, it is typically one of the first steps (the very first step being a stretch to see what sort of data you have). The decision whether or not to bin usually depends on;
1. Is my data oversampled? That is, is detail in my image smeared out over multiple pixels due to seeing or my CCD being too hi-res for my scope? If this is the case binning can convert that 'useless' resolution into useful signal gain.
2. Is my data so noisy that it holds me back from stretching out bigger detail that is not marred by noise?

Cheers,

[mcbbcn]

Hi Ivo,

Now, I realize how important is the tracking function to the whole thing. Thank you.

The luminance (AKA Ha) I shared earlier was processed using the 'tracking' function, but the RGB and RGB+Ha did not have the tracking function. The tighten function was applied to the HaRGB, not to the Luminance, should I apply tighten to both? Ha and RGB separately? or should I should apply tighten to the final HaRGB?

The image does not have Ha & luminance, only Ha. Here are the files cropped & aligned:

Red (15min): http://dl.dropbox.com/u/8971416/Challen ... ion_CR.fit
Green (15min): http://dl.dropbox.com/u/8971416/Challen ... ion_CR.fit
Blue (15min): http://dl.dropbox.com/u/8971416/Challen ... ion_CR.fit
Ha (45min): http://dl.dropbox.com/u/8971416/Challen ... ion_CR.fit


...let me know if you introduce any different steps to the HaRGB tutorial that you shared with all of us...

Thank you for the explanation on binning...and well, thanks for the explanation on everything. I'm getting all the answers I need, and I'm sure I'll have more questions but this should get me enough information to get me started with a few more processing tries, and processing learning.

Cheers,

M.

[admin]

First I create the new luminance master;
45m RGB and 75m Ha = 40% RGB, 60% Ha.

Now got my new luminance. Engage Tracking.
First Autodev indeed shows a rather noisy image. Making Autodev ignore detail under 1.5 pixels or so, pushes the noise back, while not overexposing stars.
I cancel and bin to 50%, then and redo Autodev. The noise is much less visible and the ignore detail parameter does not need tweaking.
HDR Equalise, up 'Most affected detail size' to 175 or so pixels, which works better for widefields where stars tend to visibly cluster to form super structures. Bumping up 'Most affected detail size' will make the size at when these clusters form larger, so that they don't form as quickly.

I'm going to do some wavelet sharpening;
Stars, Auto, Filter Sensitivity 10
Grow, Grow, Invert.
Sharp, Small Detail Bias 90%, Amount 400%
Mask Fuzz 3

With not much more to enhance, I stop Tracking and do final noise reduction.
Parameter [Scale 1] set to [75 %]
Parameter [Scale 2] set to [75 %]
Parameter [Scale 3] set to [50 %]
Parameter [Scale 4] set to [25 %]
Parameter [Scale 5] set to [0 %]
Parameter [Color Detail Loss] set to [50 %]
Parameter [Brightness Detail Loss] set to [40 %]
Parameter [Structural Emphasis] set to [3 pixels]
Parameter [Edge Repair Strength] set to [25 %]
Parameter [Noise Tracking Influence] set to [100 %]

Note that you could stretch the luminance manually to bring out more detail (which thanks to the binning is a possibility), but I prefer the balance as suggested by AutoDev - it gives the best noise to detail ratio.

I save my luminance file.

Now I create my RGB
LRGB: load r, g, b
Bin to 50%
Autodev.
Wipe to get rid of green
Re-do Autodev, 1.5px

Save RGB

Now I combine them;
LRGB: Load L, R, G, B
I use the following ratios (by eye) R 1.3, G 1.5, B 2.0. We'll fix colours later.

I do another wipe;
Parameter [Cap Green] set to [To Brown]
Parameter [Dark Anomaly Filter] set to [3 pixels]

Next I get rid of the colour noise;
Parameter [Scale 1] set to [100 %]
Parameter [Scale 2] set to [75 %]
Parameter [Scale 3] set to [50 %]
Parameter [Scale 4] set to [25 %]
Parameter [Scale 5] set to [0 %]
Parameter [Color Detail Loss] set to [75 %]
Parameter [Brightness Detail Loss] set to [0 %]

Modify colour balance;
--- Color
Parameter [Cap Green] set to [To Yellow]
Parameter [Blue Gamma] set to [1.00]
Parameter [Green Gamma] set to [1.00]
Parameter [Red Gamma] set to [1.00]
Parameter [Saturation] set to [165 %]
Parameter [Blue Ratio] set to [1.08]
Parameter [Green Ratio] set to [1.40]
Parameter [Red Ratio] set to [1.00]
Parameter [Mask Fuzz] set to [1.0 pixels]

Wipe sanity check;
Parameter [Cap Green] set to [To Yellow]
Parameter [Dark Anomaly Filter] set to [3 pixels]

Life module (Isolate preset)
Parameter [Saturation] set to [179 %]
Parameter [Strength] set to [50 %]

I decide I like a bit more red in my image;
Parameter [Saturation] set to [150 %]
Parameter [Red Ratio] set to [1.25]
Parameter [Mask Fuzz] set to [1.0 pixels]

Autonomous structure sharpening using the fracatal flux module;
Parameter [Filter Amount] set to [200 %]

I decide I like to drain more grain from my image;
Parameter [Green Gamma] set to [0.73]

The result;

Miquel_Heart.jpg (456.92 KiB) Viewed 20772 times

[mcbbcn]

HOLY KAO!!!!

That has to be one of the most beautiful images that I've ever seen in Astrophotography....OMG!!! How much I wish I could process like you my friend ...I know it takes practice and StarTools makes it easier than any other tools I have tried, still, I'm going to have to practice a lot to get to that level...last night I spent a quite a bit of time reading all the documentation you have on your site, and I'll continue reading...

I know you called the image Miquel_Heart...umm...I think I'm going to call it Miquel_Ivo_Heart

When I see this results, it quite humbling. First of all, I'm amazed that I could gather the photons for something as beautiful as this is, but 2nd, and this is even more amazing...C'MON!!! I'm in Portland Downtown in a Bortle 8 sky!!! not in my dreams I would ever thing I could produce an image like this with all the freaking gradients, reflections, ...what can I say? I'm shock! I think there was even a moon on that night...& yes, you pulled all the detail that I know it was in the Ha frame but I did not know how to pull...This is such beautiful amazing image.

...Thank you for processing the image & including the details about your processing. You made my day, this is quite a treat!!! ...it makes me think how many treasures are out there undiscovered by amateurs like me who give up on making their images better because they grow frustrated to the point that they think their image can't get any better...and here we go! What an example of just the opposite!!!



M.

[admin]

You should be extremely happy with your data - it's very good and a joy to work with Miquel!

[mcbbcn]

Hi Ivo,

I've process the heart nebular several times, to try to get close to your processing example and learn the process. But I've not been able to get rid of those blue & red pixels around the image, I noticed they were already in the RGB even after de-noise the color.

One more question, why don't I see red or yellow stars? I don't see them in your processing either...

Thanks for your help,

M.

[admin]

Hi Ivo,
I've process the heart nebular several times, to try to get close to your processing example and learn the process. But I've not been able to get rid of those blue & red pixels around the image, I noticed they were already in the RGB even after de-noise the color.

Great result Miquel!

You could use the 'RGB Blur' parameter for that when creating your LRGB composite. It will blur the color information. Alternatively you could put the offending pixels in a mask and 'Heal' them out.

One more question, why don't I see red or yellow stars? I don't see them in your processing either...

This has to do with how we've been removing green color hues from your image. The gradients were quite challenging and exhibited some green hues in places. Yellow has a large green component which was unfortunately also touched by the 'Cap Green' function.
I may have a play, if I get time, to see if there is a way of preserving yellow a bit better in the face of severe green gradients.