Help please - someone show me I'm rubbish at processing!

[edarter]

Hi,

Would there be anyone out there happy to have a go at processing a dataset I have of the Heart Nebula?

I'm not particularly great at image processing but I get by, however this one is giving me real trouble! It doesn't matter what I do in Photoshop I cannot seem to get anything resembling a half decent image from the stack. Especially given the integration time (about 5hrs with an astro modded Canon 600D and my 130PDS). I would post a result of my attempts at it, but to be honest I've never got far enough in to the processing to be happy enough to save it. Things go wrong pretty much from the off with initial stretches. I've tried gentle stretches, aggressive ones, colour preserving Arc hyperbolic etc etc and the end result always looks VERY red and background incredibly washed out. Any attempt at colour balance / correction or gradient removal sees a lot of the nebula just vanish. I'm stumped as to what I'm doing wrong. APP gives a bit better result, but VERY noisy and I just get nowhere with startools.

So if anyone is up for a challenge then please let me know, I would be genuinely intrigued to see what could be done with it compared to my half baked attempts!

Thanks
Ed

[teoria_del_big_bang]

There are a few people good at processing that may well give it a go, or at least tell you why your data is not processing well.
I would still persevere and at least put some image in the thread even if you do not think it is any good at all, even a calibrated, stacked and auto stretched image.

You need to find a way of letting people get to all your dataset as really need to post all lights and all calibration frames so they have the whole story. Probably too big to attach here so maybe something like Google Drive.

I am happy to have a go but I still am very much learning the art of processing 🙂 

Steve

[wimvb]

You need to supply your data if you want help. Either a link to the full dataset or to the unprocessed, stacked image. Supply basic information with the image (camera, telescope, filters used, single sub exposure time, total integration time).

[newbie alert]

Hi Ed, processing is a steep learning curve which is also teaching yourself to what you want from the image taste wise... Thats why when Flo were running the data sets you got so many varieties

You mention that it turns out very red, with the background washed out...this could be your colour balance, you can adjust the separate channels with levels to make the RGB sit on top of each other and the background sky adjusted with the black point slider but be careful not to clip your blackpoint as you will lose data.. I'm searching for a video on the heart, processing in PS.. post it as soon as I find it... Hang in there, Rome wasn't built in a day and all that

[malc-c]

Ed, as others have said, in order to advise you what you may be doing wrong you need to provide a link to the raw images (not JPEGs) so people can download, check them and then try and process them.  This way they can give you advice on what software to use in order to make the task easy for you.

You mentioned the camera and scope, but didn't mention if you have taken darks and bias exposures too.  If you have shot these, then again, place them on the same upload (google drive was mentioned, dropbox is another option.)

My experience is that processing is a dark magic artform... but I'm sure given access to the files these processing wizards will advise you best.

[edarter]

Thanks all, will post a link to dataset a bit later. Didn't want to do that if there was nobody prepared to have a go, but as everyone is advising it I will do so. I will also have another go at processing myself. 

I will also give details on darks, flats, bias etc which were all used.

Thanks

[carastro]

I'm happy to have a go, but can only open Tiff files, not fits unless that's the raw subs.

Carole 

[wimvb]

1 hour ago, carastro said:

I'm happy to have a go, but can only open Tiff files, not fits unless that's the raw subs.

Carole 

Have you seen this, Carole?

 

https://noirlab.edu/public/products/fitsliberator/

[edarter]

Hi Again,

Well below is a link to the FITS and TIFF versions of the stack. I've not had a chance to have another go at processing myself yet but hoping to do that this afternoon, will post the results afterwards.

https://1drv.ms/u/s!Avsp-INGb9mpzl3nj7lIfM6GmdEW?e=Ylg5iL

Details of the stack are as follows:

2 panel Mozaic
6 sessions spread over a few months (got to love the British weather!)
Approx 4.5hrs integration on each panel, subs were 4 minute exposures and any with guiding errors over 1arcsec were ditched immediately
Darks, flats and Bias applied during stacking in APP, stacked using Mozaic mode with LNC and MBB enabled
Camera = EOS600D astro modded on a 130PDS with a Baader MPC Mk3 CC and CLS LP clip in filter
Guiding via SW ST80 with a ASI 120MM Mini
All on an NEQ6 Pro in my back garden - Bortle 4/5 border

 

I went through all the data to remove duff subs in an effort to improve the base data before stacking, doesn't seem to have made much of a difference tbh. Criteria I used was focus, guide errors, satellite/aircraft trails, light pollutions from neighbours security lights and pretty much anything else I could spot.
Stars aren't perfect as these sessions were done before I modded the setup to reduce drawtube protrusion and improve MPCC spacing, but I've had much better results on other targets with the same setup. Main issues I'm encountering are:
-A bad gradient on the RHS of the mozaic after stretching, difficult to remove with Gradient Xterminator without killing most of the nebulosity.
-Significant shift to Red as I stretch, despite aligning the channels before starting stretches. Stretched trying normal curves method, arcsin hyperbolic with grey multiply and divide layers, both result in significant bleaching before nebulosity really starts to come out. I even tried realigning channels between stretches but that didn't work either!

As mentioned before, I'm processing in Photoshop. I do struggle with processing but can normally manage to get better results than I am getting with this dataset. APP's automated effort initially looked better but on closer inspection it was noisy as heck and I couldn't seem to reduce that without again killing the nebulosity and detail. I did also try startools, but I don't seem to get on with that at all and so while the result wasn't great that could be as much to do with me not knowing how to drive startools well enough!

Any help / pointers would be greatly appreciated as this is really denting what little confidence I had in my processing skills!

Thanks
Ed

[Laurieast]

Is this going in the right direction/what you expected? 
Colour preserving tone stretch in Affinity.
Separately adjusted the R,G,B, levels in Affinity.

 

Some loss of nebula in GraXpert, but not bad? 

Another go at the gradient.

 

 

 

 

 

 

 

Edited June 20, 2022 by Laurieast

[vlaiv]

I just gave it a quick look and data looks very nice - I think a lot can be pulled out of it.

I'll give it a go little bit later and will post the results.

[wimvb]

Quick & dirty in PixInsight:

• cropping the stacking edges
• gradient removal
• background neutralization
• stretching
• slight enhancement of local contrast

With an image like this, I would try to do star removal and process the nebula seperately from the stars, then recombine. But the image is a bit too large for my old laptop, so I left the stars untouched.

There's nothing basically wrong with this image, other than that it's still a bit noisy. More data is of course the best solution to that problem.

[ONIKKINEN]

I did 2 processes with the data, a simple 2 minute process with Siril, https://www.graxpert.com/ and ASTAP. First binned x2 in ASTAP, then gradient removal in Graxpert, then photometric colour calibration in Siril and lastly a simple stretch and a bit of a resolution reduction.

Then the one where i spent a bit of time in Photoshop with layer masks, starless processing and general fiddling around with saturation, sharpness, noise control (with and without TopazAI):

I dont image nebulae myself so im definitely not an expert in processing them, but i think your data is very nice and was quite easy to work with apart from the gradients which you struggled with too. Thankfully Graxpert makes short work of it and outputs a mostly gradient free image. I didn't do the sampler placements perfectly because the areas inside the heart nebula came out darker than the outside, so it could be improved. Below is an image of the compålex background model that Graxpert removed:

As you can see its a very complex and messy gradient and not one that can be explained by just light pollution or something normal during capture. Im going to guess you have light leaks and problems with the flats, not sure how else to explain the gradients but i would need to have a look at the flats and subs to say yay or nay more of that. But Graxpert works very nicely with flawed flats and you can salvage a very cursed image with that, give it a try!

As for why i think you struggled with the colours in the image is that the CLS filter clips most of theyellow-red areas out and really leaves only blue-bluegreen and then skips all the way to H-alpha. That makes your background noticeably blue and leaves all of the other detail as red from H-alpha. The photometric colour calibration tool in Siril attempts to create real colours but really it cannot since not all of the spectrum was captured. The result can look a bit off and that may be the issue if you try manual processing in PS. Overall i would recommend you ditch PS as soon as possible for early processing and do the core parts with something more dedicated to astrophotography, like Siril. With Siril you can create the first image in this post in a couple of clicks, no guessing fiddling required!

[wimvb]

3 minutes ago, ONIKKINEN said:

As you can see its a very complex and messy gradient and not one that can be explained by just light pollution or something normal during capture.

Imo, this isn't a gradient. In this case your gradient removal tool removed valid data as well. Careful marker positioning, as you already noted, should resolve this issue.

 

5 minutes ago, ONIKKINEN said:

As for why i think you struggled with the colours in the image is that the CLS filter clips most of theyellow-red areas out and really leaves only blue-bluegreen and then skips all the way to H-alpha.

That explains the lack of colours and the strong red signal, which I also found.

[Clarkey]

I had a very quick go at this and got the result below. Just some curves and level is Affinity.

The data was certainly not too noisy but as stated above there are some minor gradients. Also, the background is pretty light to get the red signal - as Wim says, this is probably down to the CLS filter. FWIW I live in Bortle 5/6 and I never use light pollution filters other than NB, but this is for the moon. I would rather get as much signal as possible and deal with the problems in processing. 

If you are using APP you could split the channels and then work on each separately which would probably improve the final result.

[ONIKKINEN]

43 minutes ago, wimvb said:

Imo, this isn't a gradient. In this case your gradient removal tool removed valid data as well. Careful marker positioning, as you already noted, should resolve this issue.

Yes, i have killed a not insignificant amount of the faintest nebulosity with the process. Most of the red blobs i would imagine are real signal, but the other blotches are kinda all over the place still and looks sort of familiar to my background prior to focuser upgrade, flocking and some other improvements to mechanics on my OTA, hence the gut feeling that there are issues not related to OPs processing skills.

[vlaiv]

Ok, so there are definitively some strange gradients present in the image. not sure what caused them, and it is not easy to get around those.

I binned image to improve SNR and result is sharpness and nicer looking stars at 100% (although not as much zoomed in as original)

[Budgie1]

Here's my rendition using PixInsight.

I managed to remove most of the gradients using carefully placed markers in DynamicBackgroundExtraction. There's a little left in the bottom right and I could have run DBE again, but it isn't too bad.

I binned x2 to make it more manageable for the PC and star removal makes the background much easier to work on. I didn't do much with the stars, a little saturation and one default setting in MorphologicalTransformation.

With the background, I went for detail and tried not to over do it. I left noise reduction to one of the last processes and use NoiseXTerminator set to 0.60 and 0.15 on the details. I could have gone stronger with the de-noise but wanted to leave a bit in there as it would have started to remove some of the detail. 

[wimvb]

Some of the unusual gradients may be due to the fact that this is a mosaic. If gradients aren't removed from each panel before mosaic composition, they can in effect repeat in the final image. And given the fact that this is a very difficult target for gradient removal to begin with, that may very likely be what we are all seeing.

[Budgie1]

I tried using GradXpert on this and it did remove the initial gradients, but added a lighter gradient in the bottom half of the image. DBE did a better job overall, so I stuck with that, and as I said above, maybe a second run of DBE with markers concentrated on the remaining gradients may have removed them. 

[wimvb]

I let Starnet do its thing during dinner, and processed the starless version a little harder. Then added the stars back in. In retrospect, I could have used a less stretched version for the star layer.

Here is the starless version

[edarter]

My word, all of these are much better than I have managed to achieve. Thank you all so much for taking the time to have a go at this!

What I am taking from this so far then is:

-I am indeed rubbish at processing in PS. Question about PS processing though, should I get all three curves as far left as possible individually without clipping, or should I get blue and green central in the red and bring the whole lot left until the red is as far left as possible? Also, if the green and blue peaks are narrower than red, should I stretch them to match red?

- I should lose the CLS filter and see what I get?

- Seriously look at alternative software, have been resisting the PI thing due to cost and complexity but it may be worth it given the hours I'm burning in PS getting nowhere. Also look at Siril and Affinity.

-Further mods to my setup may help. I do now have far less FT protrusion but want to flock the OTA and mount the primary on silicone 

Thanks again for the help so far!

Ed

[DaveS]

If you're looking at some dedicated software give AstroArt a go. There's a free trial, and the software itself is a lot less expensive. In my opinion it's also a bit more intuitive than PI, which did my head in, even with Warren's book.

[wimvb]

20 minutes ago, edarter said:

I am indeed rubbish at processing in PS. Question about PS processing though, should I get all three curves as far left as possible individually without clipping, or should I get blue and green central in the red and bring the whole lot left until the red is as far left as possible? Also, if the green and blue peaks are narrower than red, should I stretch them to match red?

This is what the histogram looks like for my image, after processing. Green and blue are narrower, because there is no signal to speak of, just background. Your cls filter clips most of the yellow and green. Imo, you should try imaging without the filter. Bortle 4/5 should be dark enough.

26 minutes ago, edarter said:

Seriously look at alternative software, have been resisting the PI thing due to cost and complexity but it may be worth it given the hours I'm burning in PS getting nowhere

I've seen great images processed in PI and in PS, but am not familiar with Affinity, APP or Startools. Processing is an important part of AP, and if you are willing to spend money on gear, you should, imo, also be willing to spend money on processing software. Most software allows an evaluation period. Use that to find out which software feels best.

 

[Budgie1]

5 minutes ago, wimvb said:

I've seen great images processed in PI and in PS, but am not familiar with Affinity, APP or Startools. Processing is an important part of AP, and if you are willing to spend money on gear, you should, imo, also be willing to spend money on processing software. Most software allows an evaluation period. Use that to find out which software feels best.

I couldn't agree more with Wim, but YOU have to be happy with the software you choose, both using it and with the results you're getting from it, so make use of the "Try-B4U-Buy" deals to see what suits you best.

I already had PS CS3 when I started back into astrophotography a couple of years ago, so started using that, but I soon started looking for something dedicated to AP and tried the 45 day trail of PI. After watching some tutorials (a "must" if you want to get going with PI) and following the processing I found my images were much better with PI then I could get with PS, so I bought it. Sure, it has a learning curve, but so does any new software.

When I started with PI I decided I would learn how to post-process the images and then worry about the stacking & calibration side later, I had DSS for that and was happy with the results that produced. This made life a little easier, but PI has so much in it that I don't think I've ever used half of the processes available and I'm still learning, every time I use it.

The best advise I can give, if you decide to try PI, is:

• Make use of the tutorials, others have already tried it and found a way!
• Get yourself a basic workflow
• Arrange the process icon shortcuts down the right of the screen in the rough order of your workflow
• Rename the icon shortcuts so you know what they do! On my screen "IntegerResample" is renamed "Binning", "MorphonlogicalTransformation" is "StarReduction" and "Deconvolution" is "Sharpen".
• Make sure you save the process icon shortcuts so you can reload them when you restart PI.

Whatever software you choose, don't struggle with it, this is a hobby and has to be enjoyable!