Suggested processing improvements - M42 Composite

[goldspider]

Very new StarTools user here! I've had a lot of success since I figured out how important it is to bin and crop out stacking artifacts, and now I'm looking at tweaking my workflow to wring some more quality out of my data. Here's a rough summary of my workflow:

1. Bin
2. Autodev
3. Crop
4. Develop (redo-global stretch) using "Home-In"
5. Wipe - color correct only
6. Develop (redo-global stretch) using "Home-In"
7. Wipe - light correct only
8. Develop (redo-global stretch) using "Home-In"
(Repeat 5-7 as needed)
9. Deconvolution
10. Healing (Isolate)
11. Color balance (artistic, never scientific)
12. Noise reduction

Using this on several sets of differing-length exposures, I was able to make this composite final image. I try for as "natural" a look as possible, with a light touch on color balance and saturation.

http://www.astrobin.com/full/285838/0/?real=&mod=

Though I'm pleased with how this came together, there are a few issues that seem to be cropping up in this and other sets of data I've shot recently.

-Background graininess - I could just be stretching the histogram a bit too much, but I suspect that the data is usable and can be cleaned up with some expertise in the software.

-Non-round stars - Shouldn't be the optics since I use a matched ff/fr with my scope. Guiding issue? Recently found a way to improve my OTA balancing, so I'm looking forward to seeing if that improves my tracking.

-Color balance - I suspect the combination of modified DSLR and broadband light-pollution filter is bringing in a LOT of red at the expense of blues and greens. Any suggestions for bringing these colors out more without looking artificial?

Any recommendations from the StarTools veterans here would be most appreciated!

[admin]

Hi,

Great stuff!

If you'd be happy to share the data it will be easier to assess and find possible improvements to your workflow.

Can you tell us how you stacked your data? I note that your flow does not show a HDR combine. Is that because this was performed during stacking? (that would not be advisable!)

There may be scope for simplification of the workflow, while I'm confident we can all but eliminate the noise grain.
I believe the workflow could be somewhat simplified with a single pass of the Wipe module and the use of AutoDev.

StarTools will not modify your stellar profile (unless you explicitly use the Magic or Repair functionality), so there is likely some tracking error or field flatnness problem at play.

Again, do let us know if you'd like to share the data.

[goldspider]

Here's the three sets of lights I stacked in DSS; one set of 20s, one set of 60s, and a set 300s that I blended with another night's 120s to try to get a little deeper data on the faint dust (you'll see the obvious stacking effect).

https://dl.dropboxusercontent.com/u/828 ... posite.zip

The "HDR composite" was done manually in Photoshop using layer masks per Jerry Lodriguss' tutorial.

Other than OTA balance and it's effect on guiding, what else could be throwing off the field flatness? I don't think it's my clip-in light pollution filter (which I was surprised to see, when I got it, rests fairly loosely in front of the sensor) but I wouldn't be terribly surprised if it's not sitting flush and causing minute refraction.

If the raw data would be more helpful, I'll see about finding a place to host it.

Thanks for the help!

[admin]

Hi,

Thank you for uploading the data!

In the future, try to refrain from blending multiple exposure lengths; it can cause artefacts, exacerbate noise and makes it harder for StarTools to track noise propagation.

The 20s and 300s+120s stack cover the full dynamic range - there is no need for the 60s stack (and it didn't use it in the rendition below).

I processed your image according to your expressed (colour) preferences, noting that especially M42 really warrants a more colourful approach; it is one of the most interesting areas in the night skies with a great variation of processes, chemical signatures, materials, all with their own distinct colouring. For example, the core is on of the few objects in the night skies with a predominantly green color; it is rich in O III emissions, excited by the hot blue O and B class stars there. There is brown dust, red and blue hydrogen alpha and beta emissions, as well as reflection nebulae. Of note is also the deep red star - filtered light due to dust obscurity (I believe) south of M43. It's a great keystone for colour calibration.

It's a pity a lot of AP'ers still think that it is ok (or even necessary!) to process/stretch luminance and colour at the same time, just because that was the only way possible in the olden days due to limitation of early software and/or film. These days, however we have more sophisticated tools to show colour independent of how bright something was recorded. A lot of people stil think M42's core is white - it is not, it is a teal green with a beautiful red serpentine at its southern boundary that extends into the gerater nebula.

It is unnecessary these days to use manual labor and masks to combine images of different exposure lengths (see work flow).

I processed the two stacks as follows;

(300s+120s);
--- Bin
Parameter [Scale] set to [(scale/noise reduction 50.00%)/(400.00%)/(+2.00 bits)]
--- Auto Develop
To see what we got. Nice clean data, some dark pixels here and there.
--- Crop
Parameter [X1] set to [293 pixels]
Parameter [Y1] set to [114 pixels]
Parameter [X2] set to [2372 pixels (-229)]
Parameter [Y2] set to [1627 pixels (-105)]
(Crop will remember these settings)
--- Wipe
Parameter [Dark Anomaly Filter] set to [5 pixels] (to work around the dark pixels)
--- Auto Develop
Region of Interest over Running Man + M42 complex.
Parameter [Ignore Fine Detail <] set to [3.4 pixels]
Parameter [Outside ROI Influence] set to [5 %]
--- Deconvolution
Parameter [Radius] set to [2.2 pixels]
--- Wavelet Sharpen
Default settings.
--- Color
For more adventurous renditions and finding color balances, have a look here.
Parameter [Dark Saturation] set to [4.70]
Parameter [Bright Saturation] set to [4.70]
Parameter [Saturation Amount] set to [100 %]
Parameter [Blue Bias Reduce] set to [1.08]
Parameter [Green Bias Reduce] set to [1.12]
Parameter [Red Bias Reduce] set to [1.65]
--- Wavelet De-Noise
(turning tracking off)
Parameter [Scale 5] set to [50 %]
Parameter [Color Detail Loss] set to [13 %]
Parameter [Brightness Detail Loss] set to [13 %]
Parameter [Grain Size] set to [22.6 pixels]
Parameter [Read Noise Compensation] set to [9.80 %]
Parameter [Smoothness] set to [85 %]
(save file)

(20s stack);
--- Bin
Parameter [Scale] set to [(scale/noise reduction 50.00%)/(400.00%)/(+2.00 bits)]
--- Crop
Will have remembered parameters.
--- Auto Develop
To see what we got.
--- Wipe
Parameter [Dark Anomaly Filter] set to [5 pixels]
--- Auto Develop
Region of Interest over area we want to blend in later.
Parameter [Ignore Fine Detail <] set to [3.4 pixels]
Parameter [Outside ROI Influence] set to [5 %]
--- Deconvolution
The auto mask generator by default masked out the core, as it is very close to overexposing. I masked it back in, being careful to still mask out any stars.
Parameter [Radius] set to [1.5 pixels] (being conservative here due for reason mentioned above)
--- Wavelet Sharpen
Default parameters.
--- Color
Same parameters as other stack;
Parameter [Dark Saturation] set to [4.70]
Parameter [Bright Saturation] set to [4.70]
Parameter [Saturation Amount] set to [100 %]
Parameter [Blue Bias Reduce] set to [1.08]
Parameter [Green Bias Reduce] set to [1.12]
Parameter [Red Bias Reduce] set to [1.65]
--- Wavelet De-Noise
Turning tracking off.
Parameter [Grain Size] set to [11.6 pixels]

(HDR combine)
--- Layer
(load other processed stack in foreground or background)
Parameter [Filter Type] set to [Min Distance to 1/2 Unity]
Parameter [Blend Amount] set to [50 %]
Parameter [Mask Fuzz] set to [1.0 pixels]
Parameter [Filter Kernel Radius] set to [51.0 pixels]

This yields the following image;

m42_20s_dss1_final.jpg (414.94 KiB) Viewed 6600 times

A text book, though conservatively coloured, M42/M43 and Runing Man.

Hope this helps,

[goldspider]

Thanks so much for the brilliant reference! I plan on re-stacking the 300s subs alone and going through this process again, if you think that will result in a cleaner final result. I'd suspected that StarTools might be able to automate the masking, but hadn't gotten there yet in all my reading.

I remember the very first photos of M42 I tried taking through my iPhone through my first telescope, and remember that teal green the short exposure produced. As I've dug into this hobby since then, I've always wanted a "realistic" looking end result, compared to a lot of what I see as artificial and overly-processed. Another thing I've learned is to listen to the advice of people who have been doing this far longer than I have. In that spirit, I'd love to get your thoughts on the "best" way to reproduce realistic colors, with the understanding that there is a lot of atristic subjectivity in all image processing.

Thanks again!

[admin]

I remember the very first photos of M42 I tried taking through my iPhone through my first telescope, and remember that teal green the short exposure produced. As I've dug into this hobby since then, I've always wanted a "realistic" looking end result, compared to a lot of what I see as artificial and overly-processed. Another thing I've learned is to listen to the advice of people who have been doing this far longer than I have. In that spirit, I'd love to get your thoughts on the "best" way to reproduce realistic colors, with the understanding that there is a lot of atristic subjectivity in all image processing.

You hit the nail on the head - colouring is extremely subjective! We're trying to determine the "right" colouring for something our eyes cannot detect colour in themselves. Add to that the ridiculous variations in brightness, plus the loss of a white balance reference, and any notion of "natural" colours goes out the window.

You're then left with purely academic ways of arguing a case for what a "less wrong" way of colouring might look like.

At the very least, I would personally argue that just because I decide on a particular exposure time here on earth, that things in outer space don't magically change colour (as you pointed out your iPhone picked up the teal at short exposures). I would therefore argue that it is "less correct" to show a deeper exposed M42 core white vs showing a shorter exposed M42 teal green. (for this reason StarTools implements "color constancy", undoing the colour and saturation skew introduced by the non-linear stretching performed to make detail visible)

White reference is another issue; should we use the earth's overcast daylight white reference (D65)? That doesn't sound right for celestial detail. Calibrate it against a G2V class (Sun-like) star then? Use a star field or galaxy as white reference? There is merit in arguing any of these.

Color and color perception is a very, very deep rabbit hole. However, my personal opinion is that color is too important to ignore; it yields direct and valuable clues to the viewer of your image to the make up of objects, their temperatures, what processes are going on, how they may interact with eachother, whether some may be similar to each other, how far away they are, what their history is and what their likely future will be. All deducible with a high degree of accuracy or probability from color. Distorting color ratios, or leaving color out takes away that valuable information.