NGC253, for Jase redo

[Bassnut (Fred)]

Sigh. This punk causes so much grief, but then wouldnt we all wallow in averagness without Gods critique .

Tweaked to within an inch of its life to Jases NGC253s colour balance, seemed the safest way .

The colour clipping previously (on extensive investigation) was caused by CCD stack colour create and then scaled TIFF save. Saving unscaled causes combing artifacts in PS on stretching. It sucks, so I saved each colour sub seperately as a FIT and opened with FITS liberators stretch (it cant handle RGB) in PS, then balance in PS, which is manual and much harder.

http://fredsastro.googlepages.com/home

[leon]

That is just awesome Fred, such detail and clarity.

leon

[gregbradley]

Hi Fred,

Nice detail.

To me the colour is off. The core is magentaish whereas the core actually is yellowish.

The main disk really is more brownish with yellow hues. I have seen various renditions of this galaxy but the one I like the best is R Jay Gabany's.

He tends to be heavy with saturation so that is fine if others don't agree but his 253 is one of the absolute bests. Rob Gendler has one as well and he usually is pretty accurate with his colours.

I did not know colour create in CCDstack can clip colour histograms.

There is also some Ha in the disk if you have any Ha.

Greg.

[drmorbius (Randall)]

Originally Posted by Bassnut

The colour clipping previously (on extensive investigation) was caused by CCD stack colour create and then scaled TIFF save. Saving unscaled causes combing artifacts in PS on stretching. It sucks, so I saved each colour sub seperately as a FIT and opened with FITS liberators stretch (it cant handle RGB) in PS, then balance in PS, which is manual and much harder.

I have no idea what you just said...

But it must be good because the results are fantastic... wow!

[jase (Jason)]

Are you happy with it Fred? You do realise that is all that counts. My comments are personal observations, not gospel. This is a fine tribute indeed. You've managed the colour channel gradients exceptionally well and, hey no clipping. Excellent display of detail with subtle hints of the notorious streamers that typically only mega data and careful processing can extract on NGC253. A lovely rendition to be proud of. Well done.

[Peter Ward]

Looks good to me Fred.

[AlexN]

And to think I thought it was brilliant before

Well done... Again!

Alex.

[Ric]

Lovely rework Fred, this one has a real feel of depth to it.

Top stuff

[Tandum (Robin)]

Fred,
I like the image and colours, but, what would you need to do to capture the galaxy complete as matty did here -> http://www.iceinspace.com.au/forum/a...se.php?a=46449

The object seems to be too big for your setup. I like seeing these objects floating in space where they belong.

[marc4darkskies (Marcus)]

Nice image Fred, good detail - not purple like the last one. Shame your FOV's a little too narrow though. I don't have a very good monitor here at work but I tend to agree with Greg's comments re colour.

Originally Posted by Bassnut

...The colour clipping previously (on extensive investigation) was caused by CCD stack colour create and then scaled TIFF save. ....

http://fredsastro.googlepages.com/home

I NEVER save scaled TIFs out of CCDStack - limits my ability to properly scale, balance and tweak in PS (like you've discovered), and to reduce noise (I like to do that on the unstretched lum layer).

Originally Posted by Bassnut

......Saving unscaled causes combing artifacts in PS on stretching. It sucks, so I saved each colour sub seperately as a FIT and opened with FITS liberators stretch (it cant handle RGB) in PS, then balance in PS, which is manual and much harder,

Hmmm. never noticed that - what artifacts are you talking about specifically? RGB combining in PS is for masochists Fred!!! CCDStack is best for that IMO.

Cheers, Marcus

[Jeffkop (Jeff)]

Hmmm .. looks as good a picture as the original Fred, could be even better ... but these eyes cant pic it ... Either way, its still the best picture of this object Ive ever seen

[Bassnut (Fred)]

Originally Posted by leon

That is just awesome Fred, such detail and clarity.

leon

Thanks Leon

Originally Posted by gregbradley

Hi Fred,

Nice detail.

To me the colour is off. The core is magentaish whereas the core actually is yellowish.

The main disk really is more brownish with yellow hues. I have seen various renditions of this galaxy but the one I like the best is R Jay Gabany's.

He tends to be heavy with saturation so that is fine if others don't agree but his 253 is one of the absolute bests. Rob Gendler has one as well and he usually is pretty accurate with his colours.

I did not know colour create in CCDstack can clip colour histograms.

There is also some Ha in the disk if you have any Ha.

Greg.

Yes, I found the core to be yellow too, but I pushed it to match Jases and others. Sheesh, its hard to know sometimes what to do. Its not the colour combine so much in CCD stack, its saveing with the strectch option, there are no tweaks. I tried Ha, but it was woefull, only the core showed up, so yes there is Ha there. I hadnt thought of useing that in the core, good idea.

Originally Posted by drmorbius

I have no idea what you just said...

But it must be good because the results are fantastic... wow!

Thanks Randall

Originally Posted by jase

Are you happy with it Fred? You do realise that is all that counts. My comments are personal observations, not gospel. This is a fine tribute indeed. You've managed the colour channel gradients exceptionally well and, hey no clipping. Excellent display of detail with subtle hints of the notorious streamers that typically only mega data and careful processing can extract on NGC253. A lovely rendition to be proud of. Well done.

Geez thanks Jase. The streamers are almost impossible, not even sure if I can see them here at all. I can only recall a couple of images with them. Yes Im fairly happy with it now, thanks for the prod

Originally Posted by Peter Ward

Looks good to me Fred.

Thanks Peter, a rare treat from you .

Originally Posted by AlexN

And to think I thought it was brilliant before

Well done... Again!

Alex.

Thanks Alex, unforntunately, it can always be better, but one has to stop somewhere.

Originally Posted by Ric

Lovely rework Fred, this one has a real feel of depth to it.

Top stuff

The core does seem to sort of drop in, I hope thats the way it really is

Originally Posted by Tandum

Fred,
I like the image and colours, but, what would you need to do to capture the galaxy complete as matty did here -> http://www.iceinspace.com.au/forum/a...se.php?a=46449

The object seems to be too big for your setup. I like seeing these objects floating in space where they belong.

Yes Robin, the FOV isnt good, sitting whole in space would be nicer, but I didnt want to rotate the cam with a new set of darks. The F/R is already there so I cant zoom out.

Originally Posted by marc4darkskies

Nice image Fred, good detail - not purple like the last one. Shame your FOV's a little too narrow though. I don't have a very good monitor here at work but I tend to agree with Greg's comments re colour.



I NEVER save scaled TIFs out of CCDStack - limits my ability to properly scale, balance and tweak in PS (like you've discovered), and to reduce noise (I like to do that on the unstretched lum layer).



Hmmm. never noticed that - what artifacts are you talking about specifically? RGB combining in PS is for masochists Fred!!! CCDStack is best for that IMO.

Cheers, Marcus

Marcus, loading into PS unstreched with a low count ADU medium stack needs so much stretching that "combing" artifacts appear (spiky histogram) due to just a few (can be less than 8 bits) A/D levels becoming visible.

I like moving FITS into PS via fits liberator because it has very tweakable strectch algorithms. Unfortunately it cant handle RGB FITs, so the only way is to save RGB in CCD stack as TIFFs and directly into PS. I had trouble with colour create balances with mono subs in PS so tried colour create 1st in CCD stack, which is easy. Unfortunately the scale function in CCD stack clips (it maybe ive done something wrong, but I cant see it), its now back to seperate mono FITs.

If you are importing unstreched Lum, then you must have good data. 3nm Ha especially can be dense, but low ADU count.

Originally Posted by Jeffkop

Hmmm .. looks as good a picture as the original Fred, could be even better ... but these eyes cant pic it ... Either way, its still the best picture of this object Ive ever seen

LOL, hi praise Jeff, im chuffed

[marc4darkskies (Marcus)]

Originally Posted by Bassnut

Marcus, loading into PS unstreched with a low count ADU medium stack needs so much stretching that "combing" artifacts appear (spiky histogram) due to just a few (can be less than 8 bits) A/D levels becoming visible.

Waa??? Too techno for me. If you mean that stretching low S/N data yields a noisy result (= spiky histogram) well OK.

Originally Posted by Bassnut

I like moving FITS into PS via fits liberator because it has very tweakable strectch algorithms. Unfortunately it cant handle RGB FITs, so the only way is to save RGB in CCD stack as TIFFs and directly into PS. I had trouble with colour create balances with mono subs in PS so tried colour create 1st in CCD stack, which is easy. Unfortunately the scale function in CCD stack clips (it maybe ive done something wrong, but I cant see it), its now back to seperate mono FITs.

If you are importing unstreched Lum, then you must have good data. 3nm Ha especially can be dense, but low ADU count.

I didn't think loading FITS RGB channels one at a time into PS from the same low S/N data would be noticeably better (??) I imagine you could also disrupt the colour balance you achieved in CCDStack thus making your PS exerience more challenging. Re the clipping, I'd suggest not doing any DDP or trying to adjust the gamma in CCDStack - much more control over these things in PS. I only do initial colour combination in CCDStack and save unscaled.

I import my LUM as FITS (logarithmic) into PS. Same with Ha but yes, I use a 10nm filter and my subs are usually 30 mins at the longer FL. I generally apply 2 to 4 stretching curves layers (depending on signal level). I then apply noise reduction to a masked copy of the unstretched layer. For my RGB I may apply more stretching layers and sometimes do noise reduction the same way. I typically also blur it a bit (but not always). Layer masking is a must of course.

Cheers, Marcus

[Bassnut (Fred)]

Waa??? Too techno for me. If you mean that stretching low S/N data yields a noisy result (= spiky histogram) well OK.

No, not "noisy". The spiky histogram is due to low bit depth data stretching to the point you can see each digital level. In the extreme, the image seems to have layers of stepped brightness. Stretching 32 bit files before conversion to 16 bit preserves signal level resolution ie spreads the signal over the available 16 bits.


I didn't think loading FITS RGB channels one at a time into PS from the same low S/N data would be noticeably better (??)

Well, it can be. Fits liberator cant import colour FIT files, so then it must be saved in colour TIFF to go direct in PS. If its not stretched 1st, then stretching in PS can cause the above artifacts. In this case FITs via Fits liberator pre-stretching could give a better result. I found that on saveing TIFFs in CCD stack as "scaled", clipping occured, so again, a worse result.

This is getting picy maybe, if the data is solid then the difference wouldnt be noticable.

I imagine you could also disrupt the colour balance you achieved in CCDStack thus making your PS exerience more challenging. Re the clipping, I'd suggest not doing any DDP or trying to adjust the gamma in CCDStack - much more control over these things in PS. I only do initial colour combination in CCDStack and save unscaled.

Yes, I dont do anything there either. Although many use DDP, and the advantage with it is its done in 32 bit, PS has more control with layer masks (and go-back history).

I import my LUM as FITS (logarithmic) into PS. Same with Ha but yes, I use a 10nm filter and my subs are usually 30 mins at the longer FL. I generally apply 2 to 4 stretching curves layers (depending on signal level). I then apply noise reduction to a masked copy of the unstretched layer. For my RGB I may apply more stretching layers and sometimes do noise reduction the same way. I typically also blur it a bit (but not always). Layer masking is a must of course.

Sounds like youve got that well sussed, yes I found layer masking a huge leap in image improvment.

[Garyh]

Well, I`m a bit lost in all this processing techno stuff but it sure is a nice 253 Fred!

[gregbradley]

[quote=marc4darkskies;356363]


I NEVER save scaled TIFs out of CCDStack - limits my ability to properly scale, balance and tweak in PS (like you've discovered), and to reduce noise (I like to do that on the unstretched lum layer).



Hi Marcus,

How do you open the files in PS if you don't scale the image in CCDstack?

Do you use FITs liberator?

Also I think using DDP in CCDstack - especially the auto setting, tends to clip the histogram a bit.

Greg.

[gregbradley]

Quote Bassnut:

I import my LUM as FITS (logarithmic) into PS. Same with Ha but yes, I use a 10nm filter and my subs are usually 30 mins at the longer FL. I generally apply 2 to 4 stretching curves layers (depending on signal level). I then apply noise reduction to a masked copy of the unstretched layer. For my RGB I may apply more stretching layers and sometimes do noise reduction the same way. I typically also blur it a bit (but not always). Layer masking is a must of course.


Hi Fred,

How do you do your noise reduction with a layer mask?

Greg.

[Bassnut (Fred)]

Greg. Thats a Marcus quote, I messed up the quoting thing.

[Matty P (Matt)]

Lovely rework Fred, just stunning.

Very well done.

[marc4darkskies (Marcus)]

Originally Posted by gregbradley

Hi Marcus,

How do you open the files in PS if you don't scale the image in CCDstack?

Do you use FITs liberator?

Also I think using DDP in CCDstack - especially the auto setting, tends to clip the histogram a bit.

Greg.

Yep - Fits Liberator.

Originally Posted by gregbradley

I import my LUM as FITS (logarithmic) into PS. Same with Ha but yes, I use a 10nm filter and my subs are usually 30 mins at the longer FL. I generally apply 2 to 4 stretching curves layers (depending on signal level). I then apply noise reduction to a masked copy of the unstretched layer. For my RGB I may apply more stretching layers and sometimes do noise reduction the same way. I typically also blur it a bit (but not always). Layer masking is a must of course.


Hi Fred (nay Marcus),

How do you do your noise reduction with a layer mask?

Greg.

Premise - don't stretch noise.

Of course using a layer mask is so you minimise the destruction of detail.
My recipe is pretty straightforward:

Duplicate the unstretched layer
Stretch it to taste (I use multiple curves adjustment layers)
Create a "flatten visible" layer by holding down the Alt key and selecting the menu option.
Copy this to the clipboard and delete that layer.
Zoom in on a dark area of the image to see the noise. Good to be looking at some higher S/N areas too to see the effect of the noise reduction on detail
Apply noise reduction to the duplicate unstretched layer so that all the noise is gone from the low S/N (dark) areas and most detail is retained.
Create a layer mask (hide all) to the noise reduced unstretched layer. The effect of the noise reduction will disappear
Paste the stretched image you copied earlier into the mask.
Invert the mask image. Now the whiter areas of the mask correspond to the darker (noiser) areas of the image so the noise reduction is only applied to those darker areas.
Depending on noise levels, image contrast and detail in the image, you may need to tweak the layer mask with levels or curves to make it more contrasty to maximise the filtering on the dark areas and minimise it on the lighter good S/N areas. Some blurring of the mask may also be wise
Blend the noise reduced level with the original if required.

I would've thought this is a pretty standard technique - so doesn't everyone do this?

Of course there's no substitute for good data. The effectiveness of any noise reduction technique will diminish as S/N decreases.

I find that doing this on the unstretched layer gives more latitude to fiddle the noise filtering tool in PS to best effect.

Cheers, Marcus