Dim Dust Data Chalenge! Try my NGC7023 @300mm?
Posted 08 March 2013 - 02:12 AM
I think I finally have something I am ok with - but found it VERY hard to process and wanted to inflict... er... offer the data to the forum to see how others here might manage the same data. It has often been quite instructive to see the varying interpretations and techniques!
So first off - the raw stacks as they came out of the PixInsight ImageIntegration process both in PI's native format and a more accessible 16bit TIFF:
- NGC7023-Tair3s_24x600s.fit.zip (32bit floating-point FITS file, 96M zipped)
- NGC7023-Tair3s_24x600s_16bit.tif.zip (16bit uncompressed TIFF, 37M zipped)
Next post will be my take on the data.
Posted 08 March 2013 - 02:21 AM
My two main conclusions are that I need more time and I should have run at f/6.7! The light pollution gradients did not make processing easier either so I guess a visit to a TRUE dark site would help too..
- Scope: Tair-3s camera lens (300mm f/4.5) @ F/8
- Camera: Canon XS/1000D (not modified)
- Mount: Celestron CI-700
- Guide: 50mm finder, SSAG
- Filters: none
- Software: BackyardEOS for capture, PixInsight for calibration and processing
- Exposure details: 24x 600s exposures (4 hours) @ ISO1600
High resolution version available on my site here
Posted 08 March 2013 - 03:03 AM
Posted 08 March 2013 - 11:56 AM
I need more time and I should have run at f/6.7! The light pollution gradients did not make processing easier either so I guess a visit to a TRUE dark site would help too..
Plus I think field is quite wide. Nice attempt though. Mine up close.... Thx
Nice job on your close up. I have a slightly less close up version from a couple years ago here.
However the wide field in this attempt and the "Dim Dust Data Chalenge" thread title are no coincidence. What I was trying to capture here is the outer dark nebulosity associated with the Iris.
Check out this picture google image search found me showing the extend of the.... It also better shows the smaller Ghost Nebula (barely visible in the top-right in my image). If I had known about that when I started imaging I may very well have re-framed slightly.
Posted 08 March 2013 - 11:59 AM
With regards the Tair it does do fairly well - however I several passes of star reduction here to get this result. I also seem to have an odd issue with slightly off-round stars that I *think* is in the lens but an unsure. It is not bad, but it is present.
Posted 09 March 2013 - 12:15 AM
Your data looked too interesting to pass up, so I decided to try a quick process on it.
I agree with you that it could use some more time. I pushed the data pretty hard (probably harder than I should have). Also, because I didn't want to spend too much time on it, I reduced the resolution early on. So, I lost a lot of detail potential with that.
Anyway, thanks for the processing challenge - it definitely was that!
Posted 09 March 2013 - 06:09 PM
Posted 11 March 2013 - 02:15 AM
Erik: Of course it does not match the unrealistic expectations I had in the back of my mind during capture, but ya, considering conditions and f/8 the data does at least contain the dust! Does make me wish for true-dark skies and f/4 optics though!
Posted 11 March 2013 - 02:27 AM
I also discovered part of why I had such trouble with colour calibration.... It seems during debayer PixInsight was scaling the 12bit raw data to fill the full 32bit working space (logical). HOWEVER it was doing that ONLY for the first channel (Red in this case) leaving Blue and Green "missing" data entirely in the top of the dynamic range! I simply clipped the white point on both blue and green channels till at least one pixel was 100% intensity and the starting colour balance was closer to sane then usual. Still in need of adjusting, but better then last time!
Anyway... Here is a 50% resolution (slightly cropped) version of my re-process. I am happier with this result for sure. None the less this is obviously pushing the data a bit past where it should be - just not enough signal and too much noise in the dim data. I also re-uploaded the version on my gallery so the higher-res version can be found here.
Posted 11 March 2013 - 02:29 AM
Posted 11 March 2013 - 07:26 PM
Thanks for your take Scott. What software did you use? I am not sure how you managed to keep the stars down to such a low level while still bringing dust out!
Mostly I use a combination of Photoshop CS5 and Images Plus 5.0. As far as the star sizes and the dust - there's a few techniques I use. Keep in mind that I processed this as an LRGB image, where I create a synthetic luminance from a greyscale of the RGB. That helps to minimize effects from chromatic noise.
One of the things I do is that as I stretch the dust, I ONLY stretch that range. In other words, most of my curves set a few anchor points at and below the black point, and then will set an anchor point at the brightness of the dim dust. I'll raise that part of the curve and then I'll set a couple of anchor points above the faint dust and bring them back down to the diagonal line. This makes it so that the stretching applies to the dust and minimally to the stars. When I'm all done, I end up having a curve that is basically a diagonal line with a hump around the range of the dust.
The other thing that helps bring out the dark lanes is to set points at the dark lane and at the faint dust. Then, just lower the dark lane anchor point VERY little (usually an adjustment of just 1 will make a very dramatic difference). This steepens the curve between these two points, which increases the contrast between them. Then, again, I set anchor points below this and above this to make the curve ONLY affect this range.
Still, the stars and their halos will end up growing a bit, despite my best efforts. With a field of view like this that has so many stars, the nebula and dust shows a lot better if the stars are kept to a minimum. In order to combat the growth of the stars, I'll shrink the stars at several steps as I'm stretching (i.e., stretch, stretch, shrink... repeat as needed).
For star shrinking, the very best tool I've found is IP's Special Functions | Star Size and Halo Reduction. It does a great job shrinking the stars while leaving them looking natural and not affecting the rest of the surrounding images. If you don't have IP, then Carboni's Star Shrinking action works well, as can using the minimum filter and masking it to the stars. The beauty of IP's Star shrinking is that it affects ALL of the stars - even the fainter ones that tend to start overwhelming the image as you stretch them.
Hope this helps.
Posted 11 March 2013 - 07:30 PM
Posted 16 March 2013 - 07:27 PM
Edit: Whoops, forgot to attach the image! Here it is...
Posted 19 March 2013 - 12:39 AM
If I receive permission from Sean,I will upload the full image to my Flickr account.
Permission granted! Looking forward to seeing your non-JPEG-murdered version
In fact, to make it properly official for everyone I now declare that the two stacks linked from the first post here are released under a Creative Commons Licence.
Posted 19 March 2013 - 10:33 PM
Ok, so I lost the Iris detail... but it sure was fun trying.
Posted 20 March 2013 - 06:04 PM
Ok... PLEASE share with me your colour calibration steps! The star, nebula and dust colours look much better in yours then my attempts. I know you started with the same data and use the same app so it must be my processing....
You also managed to bring out an impressive amount of detail in the Iris itself.
Posted 20 March 2013 - 06:08 PM
Francisco: Since you mentioned the Tair-3 lens a few times I figured you (or someone) might be curious so here is one of the individual subs I captured (11M .CR2 file_ to take a glance at.
Posted 20 March 2013 - 06:18 PM
I got a little carried away...no clear sky in my future anytime soon so I plan to revisit your data, it's an awesome example for learning to pull out faint stuff.
Used PS CS2, Fitswork and StarTools, got a little carried away with the noise reduction.
Posted 20 March 2013 - 07:30 PM
In any case I'll post the processing steps and parameters in PI shortly. Here are some things I can tell you off the top off my head.
The detail in Iris was done by using HDR Wavelets after applying a star mask. I did this at three scales 7, 6 and 5 with deringing turned on.
LHE improved the contrast. Saturation via the Curves tool (see Harry's tutorial) made the colors pop.
Posted 22 March 2013 - 12:01 AM
Light Pollution Gradeint removal:
DBE. Selected additional points manually. Correction applied was a subtraction.
Did a high order polymomial ABE, could have been 9th degree. Correction applied was a division. This cleared up the corners.
Did a Morphological Reduction following Rogelio's tutorial:
White Balance via the Star method and SCNR.
Atrous Wave reduction following parameters similar to Harry's tutorial. This is done with a clone applied as a mask to protect high signal areas.
Immediately with the mask in the above step still on, I did GreyCStoration Noise reduction. Here are the parameters
* Iterations 1
* Amplitude 20
* Sharpness 70
* Anisotropy 1
* Noise Scale 0.10
* Regularity 0.25
* Spatial step size 0.10
* Angular step size 5
* Precision 5
* Ineterpolation Nearest neighbor
* Fast Approximation: Checked, Coupled Channels: Checked
Histogram + Curves to stretch data
Did another star reduction. Again with a Morphological Selection as before, but milder.
HDR Wavelets. As I said before, I did this at scales 7, 6 and 5 with deringing parameter between 0.22 and 0.20
Did a 2x2 bin
Local Histogram equalization. Contrast limit was 2.5 and Kernel size about 150
ACDNR with Luminance mask, Shadows about 8, Midtones and highlights about 45. Luminance SD (1.5, Iterations 3), (Chromiance SD 5, Iterations 5)
Saturation via Curves
Unsharp Mask SD 1.5-2.0, Amount 0.60
Where I've indicated a range, it means I don't remember exact values. Do previews to check these out.
Edit: GreycStoration with the parameters I use is very computationally intensive. I have a i5-3570K at stock speeds and going away for a breather is a good idea even with this rig.
Posted 22 March 2013 - 12:18 AM