I give up. Here's 40 hours of data, can you crack this one?

[Datalord]

I've struggled with the crescent for a year. Last year I had a lot of data on it, but never achieved a result I was happy with. This year I decided to add some more O3 to improve oxygen and utilize my better processing skills.

Didn't work. I just can't get the colours right. I only have Ha and O3, but I have about 20 hours of data on each. The subs look fine to me stacked and deconvolved, but I fail.

If you try it, please post your result and what you did to achieve it. I would appreciate it immensely.

Here's the result from last year:

Here's a link to the linear stacked, DBE's, deconvolved subs:

https://www.dropbox.com/sh/swyq2bzkjggne3y/AADehllXIJjLCKAOJDhKNAKMa?dl=0

[MarkAR]

I'm struggling with my own data on this. Had a look at my latest SHO again last night and I'm still not sure.

 

 

Edited June 24, 2020 by MarkAR

[vlaiv]

27 minutes ago, Datalord said:

Here's a link to the linear stacked, DBE's, deconvolved subs:

If it's not too much to ask - could that be just linear stacked (32 bit fits / tif) without DBE and deconvolution or any other processing?

[wimvb]

1 hour ago, vlaiv said:

If it's not too much to ask - could that be just linear stacked (32 bit fits / tif) without DBE and deconvolution or any other processing?

Agree, better to get the clean Ha and Oiii masters to work on.

[Kinch]

The data appears OK - but I did not spend any time processing it......just a few minutes.

I work in PixInsight....just did a stretch (and slight noise reduction) on the two images you sent.... then put them together in PixelMath (PixInsight) as HOO : RGB.....and basically that is all I did.                              EDIT:   Actaully I also hit SCNR to remove 100% green also.

(I think that is all you wanted to see.....the data is OK and comes together fine. I am not going to process it further)

(I have reduced the size of the upload here).

Edited June 24, 2020 by Kinch
Edit in the text

[carastro]

I only just saw this post, have downloaded the files and will have a try later this evening.

Carole

[Datalord]

2 hours ago, vlaiv said:

If it's not too much to ask - could that be just linear stacked (32 bit fits / tif) without DBE and deconvolution or any other processing?

Certainly. I just uploaded those two in the same folder.

52 minutes ago, Kinch said:

The data appears OK - but I did not spend any time processing it......just a few minutes.

Thank you! I made a little blunder when it came ot the use of SCNR. I throttled it at 50% in my latest attempt, but with HOO I should just go 100%. DOH!

19 minutes ago, carastro said:

I only just saw this post, have downloaded the files and will have a try later this evening.

Thank you. I uploaded the ones right out of the stack as well.

Edited June 24, 2020 by Datalord

[carastro]

It looks very nice data to me,  

I only use photoshop:

Levels and curves on each, taken into Registar to make sure aligned.

Sorry, I had to flip it, I have a thing with some targets that I can't seem to visualise it properly unless it's a certain way up. 

Combined the channels, stretched a little more in curves, then fiddled with selective colours until I got something pleasing.

 

Edited June 25, 2020 by carastro

[michael8554]

Just to be different, I prefer the original, but with a tad more blue saturation .

Probably too "Coffee Table" for most tastes.....

Michael

Edited June 25, 2020 by michael8554

[gorann]

I also had a go at your data. I used PS and made a synthetic green by mixing Oiii and Ha 50:50. Stretching it made the Ha signal to overwhelm the blue. So I did a separate stretch of the blu channel to bring out that outer Oiii shell. Some mild NR applied to the Oiii data (Noels actions). Then I tweaked the colours using curves and selective color in PS. I also gave it a dose of HDRM in PI. No star shrinking neccessary. Like Carole I felt I needed to rotate the image 180°.

Thanks for sharing your data - defenitively nothing wrong with it and great fun especially since there is no astrodarkness here until late August!

EDIT: I added a second version with more Ha in the background.

Edited June 25, 2020 by gorann

[Datalord]

2 hours ago, michael8554 said:

Just to be different, I prefer the original, but with a tad more blue saturation .

Probably too "Coffee Table" for most tastes.....

It's a good attempt, but yes, I think the reason I disliked my version was the "Coffee Table". 🙂

12 hours ago, carastro said:

Combined the channels in layers, stretched a little more in curves, then fiddled with selective colours until I got something pleasing.

I think this is more what I had in mind in the beginning. Thanks!

1 hour ago, gorann said:

I used PS and made a synthetic green by mixing Oiii and Ha 50:50. Stretching it made the Ha signal to overwhelm the blue. So I did a separate stretch of the blu channel to bring out that outer Oiii shell.

I really, really like how you made the blue shell visible on both top and bottom. I knew the signal was present in the O3 layer, but I never managed to tease it out. Tried all sorts of trickery with linear fit and masked stretching, but I never managed to get this result. Thanks!

[vlaiv]

I have not forgot about this thread, but I did run into difficulties for some reason.

Starnet++ is not doing very good job on star removal with this data. Although on this scale it looks nice (still linear, but stars removed so much more nebulosity is seen):

there are quite a few artifacts once zoomed in - which currently prevents me from doing decent processing - like these:

[Datalord]

39 minutes ago, vlaiv said:

there are quite a few artifacts once zoomed in - which currently prevents me from doing decent processing - like these

Ah, those are from starnet++. You can solve those by reducing the mesh, say from 128 to 32, but processing time increases. I usually use starnet to make my starmasks and use 32. On my i7-9900K processor it takes about 15 minutes...

[vlaiv]

1 minute ago, Datalord said:

Ah, those are from starnet++. You can solve those by reducing the mesh, say from 128 to 32, but processing time increases. I usually use starnet to make my starmasks and use 32. On my i7-9900K processor it takes about 15 minutes...

I was thinking of messing with mesh size - that is obviously way to go.

Regardless, I might need to touch it up by hand as it mistakes some bright nebula patches for stars and removes those as well.

[gorann]

This is really not an easy Ha Oiii object to process (but nothing wrong with the data). When boosting the the Oiii signal that covers the whole nebula it makes the red Ha signal of the nebula look a bit dull, but mabe that is how it should be. In any case, here is a new version (selective colour tweak in PS) where I think the nebula stands out a bit more.

 

Edited June 26, 2020 by gorann

[Datalord]

I really like the background. You managed to use the data present in it. I definitely have to give this another try myself now that I see what is possible.

[MarkAR]

I think the Crescent causes a lot of problems because there is almost too much data and the NB's overlap a lot.

 

[gorann]

8 hours ago, MarkAR said:

I think the Crescent causes a lot of problems because there is almost too much data and the NB's overlap a lot.

 

Yes Mark, the Oiii glow from the shell obviously covers the Ha inside and obscures it. A version with less Oiii should probably make the central structures stand out more clearly. So I did the experiment. I processed a version without the boosted Oiii signal (first image below) and then added this as a layer to my previous version in blend mode Lighten in PS (second image below). I like it as the central structures stand out a bit more.

[Kinch]

For anyone interested - here is one that I reworked very recently.....the data is from 2018.

https://www.astrobin.com/mi11q6/?nc=user

 

[MarkAR]

Very nice Kinch.

 

[astro_john_palmer]

Hi, thanks for sharing your data. I have never worked with combining narrowband images before so it was great practice for me for when I eventually get a mono setup.

I probably went a little bit too heavy with the stylisation or as Michael put it - "Coffee Table" - but I love the way the crescent looks like a fire raging in space, so I went with a flame like golden orange for the crescent with the blue colours I think complimenting that nicely. I left the background HA a little more traditional red.

Sorry I did not document how I got here - there was quite a lot of trial and error.

[MarsG76]

I think that your crescent looks fine.. but to adjust colors, try creating a mono luminance layer over a slightly defocused (gaussian blurred) color data, and adjust the color data either in color balance or hue adjustment....

 

 

[gorann]

On 28/06/2020 at 20:46, astro_john_palmer said:

Hi, thanks for sharing your data. I have never worked with combining narrowband images before so it was great practice for me for when I eventually get a mono setup.

I probably went a little bit too heavy with the stylisation or as Michael put it - "Coffee Table" - but I love the way the crescent looks like a fire raging in space, so I went with a flame like golden orange for the crescent with the blue colours I think complimenting that nicely. I left the background HA a little more traditional red.

Sorry I did not document how I got here - there was quite a lot of trial and error.

Nice nebula but never forget to watch out for what you do to the stars while processing. Protect them with masks when you stretch the data. Ideally their cores should not be bigger (blown out) than in the unstretched image.

Edited June 29, 2020 by gorann

[vineyard]

Hello, thanks for sharing this.  I had a hack at it too (complete newbie to processing mono so this is my first ever mono attempt - am thinking of making the switch so I thought it would be interesting to see what is involved!).  A lovely set of mono images you took there.  Anyway, here's my clumsy attempt (downsampled).

 

[ultranova]

Cracking image DataLord, well done,

I think Beauty is going to be in the eye of the beholder on this one.

Paul