Dark halos after defringe star astroimages

Software RawTherapee
1

I LOVE how RT mitigates lens aberration! But I am left with dark halos after hours of fiddling with settings. I have tried all demosaic choices with a range of parameters… maybe I missed one? I hoped that impulse noise reduction would help, but no. Any settings/workflow to get rid of those? It would be great if those could be filled with nearest neighbor.
Thanks!
Neutral on the left; sorry not quite same zoom…

image
image1456×504 150 KB

2

Are you willing to upload this RAW so the RawTherapee crowd can have a look? Bit hard to give advise based on the above jpeg and not knowing what it is you have(n’t) done.

No problem if you want to keep the RAW to yourself, I’d understand :slight_smile:

3

When I do astro shots Alister, I find sharpening is usually the culprit so I tend to process with little or no sharpening, then sharpen the output from behind a mask in Photoshop afterwards.

PS nice round stars, good tracker setup!

4

Thanks folks, your wisdom/experienced is much appreciated.

Here is a folder holding a Canon CR2 and the associated pp3; should have done that off the bat, apologies: I did not do any sharpening at this stage (being aware that deconvolving often produces dark ring halos, though I’ve never done that).
https://drive.google.com/drive/folders/1BWnJOtBXY-StsWkmETWzpFi-LhoEa238?usp=sharing

It was only after I got home that I realized that it was at f/4.5 instead of the 6.3 I usually shoot at, so there is a fair bit of color. I’m new to defringing, so hard to tell how strong to go.

Since you’re here , are there RT methods to disaberrate (ooh, a new verb!) the radial stretch? Geometry seems to do a bit of it, but may not be the right tool.

Persnickety folk will notice that the aberrations are not uniform. I measured the tilt of the lens using Samuel Chia’s running track method, and just today shimmed it, and I get to test it tonight, first clear night in a month, no joke; worst cloud on record. Tilt is one hell of a rabbit hole.

5

Hi Alister,

Had a look at your image and the first thing I noticed: The lens used is not supported.

Both RT and DT pick the wrong one (on auto setting). You opted for the same one that DT choose, but RT warns you about this one. I suspect this to be at the core of this issue. For example: Choosing None as lens profile gives you a slightly better result (just looking at the blackness around the star) and experimenting a bit shows that the influence per lens varies, significantly at times.

Using and based on the (lens) setting you choose:

I did not find a fully satisfactory solution to this problem, but playing with both defringe and haze removal might get you a slightly better result. I’m not at all familiar with what else you need to do with this image (I’m assuming this is part of a stack) so the haze removal option might have unwanted side effects for the rest of the processing.

Comparisons @ 500% zoom:

003.org.500
003.org.5002048×1152 730 KB

003.alt.500
003.alt.5002048×1152 847 KB

@ 100% zoom:

003.org
003.org2048×1152 1.99 MB

003.alt
003.alt2048×1152 2.04 MB

You might want to give it a try and re-adjust the settings shown (I just looked at the problem spot and am not at all familiar with astro photography and its ins and outs).

Just in case: alternate sidecar

You mention the Lens / Geometry module, which is the correct tool (in Camera-Based mode). But I do believe it is based on the, in this case incorrect, info found in the Lens Profile module…

EDIT

I started with a clean slate and am able to mitigate this even more:

003.fs
003.fs2048×1152 720 KB

003.fs.pp3 (16.2 KB)

A different demosaicing and, mainly, different settings for the Noise reduction module (basically the median filter part). Everything you do with denoising comes at a cost, but maybe this is acceptable and you can build on it.

6

A slightly different approach than that from @Jade_NL : no lens data used here. Just standard RT tools.

I think in astrophotography the color of the halos of stars are important to identify them, but as I have no idea what are color artifacts or proper star halos, I’ve made my best guess here.

One thing to note is that your shared image (img_0003) is not the same than that from your original screenshot (img_0006), but it’s close enough, I think.

So here is what I can get, enhancing a bit the overall contrast (with the Exposure tool, Black slider), but (hopefully) respecting the stars colors:

img0003
img0003771×539 83 KB

IMG_0003.jpg.out.pp3 (14.1 KB)

Maybe you can work with it as a starting point.

1 Like
7

@skygaze You have some pretty uneven lens distortion, especially near the lower right corner. Are you sure your setup is all right?

8

Thanks for your rapid input folks! The supported lens, 300mm f/4 IS USM, is there when I search down the Canon list. it is just not automatically picked up. It does indeed raise a flag about crop factor used for lens profiling larger than that for current camera. Which is full frame. Ah I suppose that means that no one entered the data for a full frame camera. Another to-do… sigh.
Indeed @Jade_NL Jade_N, there is likely a trade-off between de-fringing and star color. A legitimate small red or blue star might get defringed into white. I am too new at this to have experience. I will try the non-lens specific approach.
@XavAL , the shots are different but they are the “same”, in that the camera is on a tracker, so only pixel level differences, noise, sky condition are different. That said, I do not play with exposure/contrast here, that occurs after I stack multiple shots (in Siril), then postprocess and noise reduce with StarTools (all this to avoid Photoshop!). That said, I will download the pp3 and have a look.
And @Thanatomanic, you are correct, there is definite unevenness, which is why I went down the tilted-lens rabbit hole. I’m only now starting to suspect that I may have received this lens as a gift because the previous owner dropped it.

9

Then you may wish to completely avoid the Impulse Noise Reduction tool, as it removes the tiniest stars, as well as creates artifacts in bigger stars (darker pixels in the center of the star).

In my pp3 the Noise Reduction tool is active with default values, meaning that it won’t remove any noise but color noise.

If you definitely follow the defringing path, you may have to learn quite a bit how that tool works in RT, as it is wonderful, but also very picky.

If I may, I would recomment that you read the Spanish RawPedia page about Defringing. With a good translator and playing with RT at the same time, you shouldn’t have many problems understanding it.

And thinking about your astro images problem, perhaps you should pay special attention at the third practical example and understand what’s happening in full. If not, you won’t get good results.

Be aware that this tool is like a double-edged sword, so there will be times where you won’t get good results, no matter what.

HTH

2 Likes
10

Someone fluent in Spanish and English should translate those. This is the second time that the Spanish documentation has a much more detailed section then the English one (Colour Toning being the other one I know about). I don’t mind doing the DeepL thing, but technical stuff does get lost in translation…

Anyway: Thanks @XavAL! Keep pointing these out, it is appreciated (at least by one person :smile: ).

1 Like
11

:blush:

This is going to be clearly OOT, but there you go!:

Sadly, I know people don’t like to read a lot :sweat_smile:

3 Likes
12

Cool and bookmarked.

I’m guessing I already know a lot about most of these, but some in-and-outs elude me. So it’s nice to have a fallback if the main/English doc lacks the info.

OK, back on topic from here out on…

13

I just happen to have a first-language Spanish astro friend in town, so I am in luck. Much thanks for the link!

1 Like
14

Google translate did a pretty good job with it. Definitely enough to follow along.
https://rawpedia-rawtherapee-com.translate.goog/Defringe/es?_x_tr_sl=auto&_x_tr_tl=en&_x_tr_hl=en-US

1 Like
15

I must admit I’m quite surprised how good is the Google translation on this page. I’m not certain if everything is grammatically correct, but overall it’s quite spot on (from my point of view).

However, as @Jade_NL said, some technical terms are not really correct:

  • purple border removal: should be defringe/defringing
  • purple edge removal: should be defringe/defringing
  • purple border: fringe
  • where it says the optics: it should be read the lens
  • in the third example 3. Although: it’s better translated as 3. However

Aside from that, I think you may get a good understanding about how the tool works straight away.

16

I agree with @XavAL. I remember doing online translation of few spanish pages of rawpedia with google translate thru firefox. It was pretty good.

Edit: The FF add-on was “Translate Web Pages”