Wrestling with ngc1333 first lrgb image

[Allinthehead]

Hi all. This is as the title says my first effort at an lrgb image using the zwo asi1600. 

L=10 hrs of 180 second subs Gain 164

RGB=9 hrs combined of 180 second subs Gain 260

I used dss to stack the images which were taken over multiple nights(5 i think). I used the best rated luminance sub as a reference frame for stacking and aligning the rgb. Then processed in cs6 luminance separately. Combined the rgb channels, used gradient exterminator, stretched then pasted the luminance over the rgb as a layer and changed the mode to luminosity.

I've tried adjusting the saturation, brightness and contrast but this is the best i can get out of the data. For the time i put into it, i'm disappointed. Can anyone spot any errors in my workflow or suggest anything that might help.

Thanks for looking.

Richard. 

[Ewan]

I don't know too much about using such short subs on what appears to look like quite a dim target.

Unless the target is very bright I would feel that subs longer than 5 mins each would be necessary, I don't know that much as I use the 460EX sorry.

[johnfosteruk]

I hope you don't mind Richard, I've been practicing processing with various pilfered data recently in preparation for getting a decent mount so I had a play with your image.

Although my effort is a little noisy there's plenty in there and someone more experienced than I would get more out of it I'm sure.

I brought the dust clouds out with a little fine curves adjustment, pinning the background and taking it down 2 points and the dusty stuff up 2 points.

Then converted to lab colour, increased the contrast in the a and b channels to 40 to increase saturation without increasing the noise.

Quite a bit of noise though at this point so 2 duplicate layers, cloned out the brightest points on the middle layer, gaussian blur 250px, set the top layer to lightness to knock it back but quite a lot of detail lost at this point so upped the black point of the blurred layer in levels by 7 points to compromise. Quick mild noise reduction filter too for good measure.

Then finally to really bring out the detail I did Paddy's trick with a blurred saturated mask. Duplicate layer, mode set to colour, blurred 15px, increased vibrance and saturation to about 40 each. Add a mask and paint in the stuff I wanted to enhance.

flatten and 50,3,1 unsharp mask.

Don't know if it is an improvement or not as it's still noisy now but it's my two penneth. 

[johnfosteruk]

5 minutes ago, Ewan said:

I don't know too much about using such short subs on what appears to look like quite a dim target.

Unless the target is very bright I would feel that subs longer than 5 mins each would be necessary, I don't know that much as I use the 460EX sorry.

Doesn't the asi1600 and it's brethren require much shorter subs though (see Gina's thread). Seems like ZWO are rewriting the rulebook.

[MikeODay]

Lovely image Richard - it is well worth clicking on the image to see the large version; it is wonderful.  I love the depth and detail in the dark nebula.

I like John's tweaked version too, although I think the original wide field version gives the nebula and dust clouds a nice context.

[Allinthehead]

18 hours ago, johnfosteruk said:

 

I hope you don't mind Richard, I've been practicing processing with various pilfered data recently in preparation for getting a decent mount so I had a play with your image.

 

Don't mind at all you did a nice job on the background better than i did. What's Paddys trick?

19 hours ago, Ewan said:

 

I don't know too much about using such short subs on what appears to look like quite a dim target.

 

I think you could be right however this camera is designed for short subs. I'm starting to think it could be a problem with the rgb as i had to use a heavy dose of gradient exterminator. I wonder would the fact that i used a gain of 260 for rgb vs 164 for lum be a factor?

 

19 hours ago, mike005 said:

 

Lovely image Richard - it is well worth clicking on the image to see the large version; it is wonderful.  I love the depth and detail in the dark nebula.

 

Thanks i think there could be a decent image in there. If i get time tonight i'll try again.

I appreciate all the input thanks all.

[ollypenrice]

This is one of the hardest targets I've ever imaged so I think it might not have been the perfect choice to help you learn a new camera!

Johnfosteruk, you made some good moves but it is vital not to apply either sharpening or noise reduction to the whole image. If you do this you'll be pulling the data this way and that and damaging it. NR needs to go only into the faint signal and sharpening only into the bright, but with the stars excluded.

If Richard would like to link us to his unprocessed L and RGB TIFF files I'm sure that lots of us would be very intrigued to process data from this new technology. I certainly would!

Olly

[johnfosteruk]

35 minutes ago, ollypenrice said:

Johnfosteruk, you made some good moves but it is vital not to apply either sharpening or noise reduction to the whole image.

Yup, I know I should have used a star mask but as it was not the unprocessed I was having a play really to see if there was anything more there rather than a serious effort. I like you would love to have a go at the unprocessed data (I had a go at yours Olly but didn't feel my efforts worthy of publication )

2 hours ago, Allinthehead said:

What's Paddys trick?

It's the process I detailed in that paragraph Richard - Duplicate layer, mode set to colour, blurred 15px, increased vibrance and saturation (of the duplicated, blurred layer) to about 40 each. Add a mask and paint in the stuff I wanted to enhance (paint white onto the black mask).

[Allinthehead]

I think this is an improvement. I'll put the data up on dropbox.

 

 

 

[Allinthehead]

Here are the files. One caveat is that i have since learned what i suspected that i used too high a gain setting for the lum and also the rgb. Olly i agree it was an ambitious target for my first lrgb image.

 

https://www.dropbox.com/s/albrmialcvslunr/newestngc1333.TIF?dl=0

https://www.dropbox.com/s/fbkm29btqxk4isb/Ngc1333red95align.TIF?dl=0

https://www.dropbox.com/s/dyfomj21uuf31wa/ngc1333green90align.TIF?dl=0

https://www.dropbox.com/s/0dwac2v23t0ul2i/ngc1333bluealign93.TIF?dl=0

 

 

[MikeODay]

46 minutes ago, Allinthehead said:

I think this is an improvement. I'll put the data up on dropbox.

That's looking really good.

[Ikonnikov]

Looks good, dust is very nice. The reflection nebula part is a more intense blue colour in other images I've seen but I suspect it just needs more data to allow you to bring it out (even with the new CMOS technology I guess this this object always want more subs!).

Paul

[Allinthehead]

47 minutes ago, mike005 said:

That's looking really good.

Thanks Mike

47 minutes ago, Ikonnikov said:

Looks good, dust is very nice. The reflection nebula part is a more intense blue colour in other images I've seen but I suspect it just needs more data to allow you to bring it out (even with the new CMOS technology I guess this this object always want more subs!).

Paul

Thanks Paul getting colour out of the image has been an issue for me. More data would always help i suppose.

[PatrickGilliland]

I actually think at 180 second subs this was a good result.  Yes it can be 'processed' more but you are limited by the data you have.  600 sec L frames will make world of difference and 300 RGB would be good.  Evaluate your final image based on the ingredients used, nice result I would say.

Paddy

[Allinthehead]

6 hours ago, PatrickGilliland said:

I actually think at 180 second subs this was a good result.  Yes it can be 'processed' more but you are limited by the data you have.  600 sec L frames will make world of difference and 300 RGB would be good.  Evaluate your final image based on the ingredients used, nice result I would say.

Paddy

Ya you're right. Looking at the second version i did i'm reasonably happy. I've a lot to learn on the capture and processing side of things, and given it's my first lrgb i should be less critical. Easier target next time i think but i just couldn't resist this one after seeing a few fabulous images of it recently on the forum.

[PatrickGilliland]

Progressing with every version.  Just keep at it if you keep improving with every version looks like a super training target.

I liked the colours and softness of the first version - i see where it going with the extra definition etc but consider that you can only push the data so far and it will breakdown no matter how well you process.

Paddy

[Allinthehead]

1 hour ago, PatrickGilliland said:

Progressing with every version.  Just keep at it if you keep improving with every version looks like a super training target.

I liked the colours and softness of the first version - i see where it going with the extra definition etc but consider that you can only push the data so far and it will breakdown no matter how well you process.

Paddy

Thanks for the feedback. I'll try again next year with the Star 71

[wimvb]

Here's my attempt with your data. I used it to experiment with various new (for me) techniques in PixInsight, so I think you can spot some errors.

I suspect that you didn't use any flat calibration, as there was a strong gradient in the images you supplied.

I had to use DBE twice, once to correct vignetting, and once to remove a gradient.

Rough workflow:

rgb combination

Then rgb + lum were processed roughly the same:

dbe x 2

colour calibration

deconvolution on stars

masked stretch

tgv noise reduction

rgb colour saturation

lum exponential transformation to get more definition in the dust (with inverted luminance mask)

rgb: replaced the L in L*a*b mode with the luminance data.

Downsampled

[Allinthehead]

Nice job wim.

[ollypenrice]

BTW, which incarnation of the Star 71 is this, the original or the new 4 element?

Olly

 

[Allinthehead]

34 minutes ago, ollypenrice said:

BTW, which incarnation of the Star 71 is this, the original or the new 4 element?

Olly

 

Hi Olly. I got the new 4 element checked and tuned from Flo. The above image was captured with the ed80.

Richard.

[ollypenrice]

22 minutes ago, Allinthehead said:

Hi Olly. I got the new 4 element checked and tuned from Flo. The above image was captured with the ed80.

Richard.

I don't know if you saw his request but Gnomus is keen to have a look at data from the new version.

Olly

[Allinthehead]

27 minutes ago, ollypenrice said:

I don't know if you saw his request but Gnomus is keen to have a look at data from the new version.

Olly

I didn't see it but i'll send him a pm.

[AKB]

@Allinthehead,

This is a super set of data to play with... thanks for posting.  I am a complete neophyte at this sort of thing, but I wanted to have a go. Somewhat hampered, perhaps, by not having PI or similar tools, hence no DBE, but do what I can with Nebulosity and a smattering of 'pixel math' using Matlab (it's what I know.)

My processing chain is along the lines of:

• bin x2 and crop (I'm not used to big images)
• crude simulation of DBE (perhaps - does anyone actually KNOW how that does what it does?)
• a single iteration separately on L R G B of an edge-preserving multi-scale decomposition filter (in Matlab)
• LRGB synthesis
• a bit of stretching

There's no deconvolution or manually prepared masks.  I'm obviously struggling with the colours, per the OP, so any further advice appreciated.  But, to me, the result is at least OK. 

Apologies to all if this takes the thread off-topic.

 

[simmo39]

I have been bitten by this target to, not an easy target to get right. After doing about 10 hours or so with my DLSR im still no where near your fantastic image.