Bicolor Crescent (NGC 6888)

[Rodd]

I finally finished the bicolor Crescent I have been working on for a few weeks now.  I mistakenly captured about 6 hours of SII data due to mislabeling filter positions.  I thought that 5 hours of SII would come in handy, but I could not find a use for it.  I tried a true HST pallet, but the image was too purple for me.  I also tried to assign the SII to the green channel, but that seemed to take away saturation and contrast from the blue envelope.  So I left the SII out, and assigned OIII to both blue and green.   I'm feeling very anticlimactic right about now.  I really like the way my Ha and OIII stacks came out and was expecting a solid image.  Not the case, I'm afraid (at least from my point of view).  The image was fully calibrated and processed in Pixinsight (Bacground extraction, Background Neut, Color Calibration, Curves for saturation an contrast, a bit of denoising, and a tgouch of sharpening)

Ha:    20  20minute subs

OIII:   21 30min subs

Flats: 33 Ha, 33 OII

Bias: 400

Darks: 33

Attached are 2 versions of the final image.  I am sure that many will suggest that the image needs allot more data.  This could be true, though I have seen much better from less data. I would like to insert RGB stars however, I'm not sure its worth it.  Anyone have any ideas about                                                

 

[nucdoc]

Of the two images you posted I really like the one on the right, (more purple and blue), it is quite striking !

 

The first one looks too green.

 

I have applied SCNR noise reduction to remove the green and posted below, hope you do not mind

 

 

[KyleStoke]

Hi Rodd, What do you feel is wrong with the image?

[Rodd]

5 hours ago, KyleStoke said:

Hi Rodd, What do you feel is wrong with the image?

Kyle,

The easiest way for me to describe it is...my image looks like crap compared to the beautiful image I found online that was shot with the same equipment and equal or shorter exposure times.  Beyond that it is hard to say.  If one has to convince themselves that an image is good--there is something off.  Processing through the linear state is straightforward.  Once the image is non linear the first curves adjustment is straightforward.  By then, good data should be looking pretty decent as far as color balance and overall appearance.  After that, one turns a nice image into a superlative image.  Still working on nice......but I have run out of options (and patience)

[Rodd]

Actually-let me rephrase it.  The image just plain sucks.  take a look at some of the other images on this forum and you will see why.  At this point I am completely fed up.  No idea what to do.  Guiding, focus, seeing, transparency were all very good.  I collected almost 20 hours worth of data. I dithered and I calibrated. I will say for the umpteenth time--I am missing a key element.  I find it strange (it would be quite interesting if it weren't related to something causing me so much aggravation) that something so obvious to me is not observed more readily by those individuals whom produce the work to which  I aspire.

[KyleStoke]

3 hours ago, Rodd said:

Kyle,

The easiest way for me to describe it is...my image looks like crap compared to the beautiful image I found online that was shot with the same equipment and equal or shorter exposure times.  Beyond that it is hard to say.  If one has to convince themselves that an image is good--there is something off.  Processing through the linear state is straightforward.  Once the image is non linear the first curves adjustment is straightforward.  By then, good data should be looking pretty decent as far as color balance and overall appearance.  After that, one turns a nice image into a superlative image.  Still working on nice......but I have run out of options (and patience)

I guess we are all our own harshest critics.

To my eyes and I say this looking on a laptop screen that I don't fully trust. The image perhaps looks a bit soft maybe a focus/resolution issue or maybe just my eyeballs.

[gnomus]

1 hour ago, Rodd said:

Actually-let me rephrase it.  The image just plain sucks.  take a look at some of the other images on this forum and you will see why.  At this point I am completely fed up.  No idea what to do.  Guiding, focus, seeing, transparency were all very good.  I collected almost 20 hours worth of data. I dithered and I calibrated. I will say for the umpteenth time--I am missing a key element.  I find it strange (it would be quite interesting if it weren't related to something causing me so much aggravation) that something so obvious to me is not observed more readily by those individuals whom produce the work to which  I aspire.

Rodd

You are being a bit hard on yourself.  Take a step back and think about what you've been able to do here.  Isn't it incredible that you can produce an image of this object using amateur equipment from your backyard?  You have managed to capture that faint blue shell around the crescent.  I think the image is pretty good.  And I think you are very close to greatness.

The hardest thing I found when starting out in this game was knowing when to stop.  Each 'tweak' seemed to reveal a little bit more nebulosity.  Each 'tweak' seemed to be a slight improvement.  However, I would stand back after an hour or so and realise that I had somehow managed to 'lose' the image.  All that extra processing had cost me something.  I still struggle not to over-process images.  With the NB filters especially, I found that I could get quite good Ha, but the other filters were much less impressive.  I would try to get the same detail out of those filters as I had out of my Ha.  I got very noisy pictures and my stars were too loud and a bit bloaty (just like me).

I downloaded your JPEGs.  Here is the photoshop histogram for your OIII.  

      

I think this suggests that you have overstretched your images a tad.  There is a gap to the left.  Using the eyedropper tool in PS I got background values in the mid 50s to low 60s for both Ha and OIII.  People who know what they are doing (and I ain't one of them) tend to aim for the low 20s and I think their 'peaks' would be a little more to the left too.  The problem with going for too much stretching is that there comes a point at which you get an eruption of noise - I got this a lot when I tried to stretch my OIII to match my Ha.  

If I were you, I'd go back to the original linear stacks and do less stretching and less processing.  PixInsight is a great tool but I find that it is very easy to end up overdoing things in it - and it has no mechanism for 'dialling back' like PS does.  Over-saturating can lead to noise too.  I'd try to get a bit more control of stars too.  If you want to stick to a PI only approach, have a look at masked stretching - make copies of your linear images and do a histogram stretch on one and a masked stretch on the other - see which you prefer.  You can use Morphological Transformation in conjunction with Star Masks to dial back the brightness of the stars a little.  Be judicious with sharpening - use masks and avoid sharpening stars and background.

You may find that a little less produces a way lot more.  

Good luck

Steve      

 

[Rodd]

11 minutes ago, gnomus said:

Rodd

You are being a bit hard on yourself.  Take a step back and think about what you've been able to do here.  Isn't it incredible that you can produce an image of this object using amateur equipment from your backyard?  You have managed to capture that faint blue shell around the crescent.  I think the image is pretty good.  And I think you are very close to greatness.

The hardest thing I found when starting out in this game was knowing when to stop.  Each 'tweak' seemed to reveal a little bit more nebulosity.  Each 'tweak' seemed to be a slight improvement.  However, I would stand back after an hour or so and realise that I had somehow managed to 'lose' the image.  All that extra processing had cost me something.  I still struggle not to over-process images.  With the NB filters especially, I found that I could get quite good Ha, but the other filters were much less impressive.  I would try to get the same detail out of those filters as I had out of my Ha.  I got very noisy pictures and my stars were too loud and a bit bloaty (just like me).

I downloaded your JPEGs.  Here is the photoshop histogram for your OIII.  

      

I think this suggests that you have overstretched your images a tad.  There is a gap to the left.  Using the eyedropper tool in PS I got background values in the mid 50s to low 60s for both Ha and OIII.  People who know what they are doing (and I ain't one of them) tend to aim for the low 20s and I think their 'peaks' would be a little more to the left too.  The problem with going for too much stretching is that there comes a point at which you get an eruption of noise - I got this a lot when I tried to stretch my OIII to match my Ha.  

If I were you, I'd go back to the original linear stacks and do less stretching and less processing.  PixInsight is a great tool but I find that it is very easy to end up overdoing things in it - and it has no mechanism for 'dialling back' like PS does.  Over-saturating can lead to noise too.  I'd try to get a bit more control of stars too.  If you want to stick to a PI only approach, have a look at masked stretching - make copies of your linear images and do a histogram stretch on one and a masked stretch on the other - see which you prefer.  You can use Morphological Transformation in conjunction with Star Masks to dial back the brightness of the stars a little.  Be judicious with sharpening - use masks and avoid sharpening stars and background.

You may find that a little less produces a way lot more.  

Good luck

Steve      

 

Wow--thanks Steve.  I appreciate the response and agree with everything you say except for one glaring thing (can you guess?).  Something about greatness.  anyway--I agree completely with the assessment.  Also, while PI may be "the thing"  I am starting to get the feeling that PS is easier and from the images I've seen--better--or just as good.  I am committed to PI for now. I can't tolerate another learning curve.  But it is damned hard to use.  Star masks are very complicated, and many of the approaches you mention are difficult.  I've attached my basic image just out of the non linear state, prior to any curves or saturation tweeks.  I also attached what I think is a better version, though perhaps over processed.

[gnomus]

Hi Rodd

I think I prefer the first of the two you posted.  The second one is too "full on" for my taste and, in my view,  you are giving up some of the detail in the Ha clouds because all the reds have been turned up to '11'.   How about something in-between (I hope you don't mind me tweaking)?

 

The image I downloaded still had a black point in the 50s and was excessively noisy.  I think with the amount of data you captured you should be able to get a cleaner image if you stretch it less.  I'd still counsel going back to the un-stretched stack and being a little less heavy handed.  The above is just 10 mins or so in Photoshop. 

   

[ollypenrice]

Pixinsight seems to tempt people into roaring mania. No, softly softly. Your first reprocess stops a tad short of what the data has to give. It looks great. The very colour saturated ones take it way too far.

I think your gentle image - the one Steve liked - is lovely. I'd give it a slight boost in saturation (slight!!!!!) and bring the greens a tad towards blue. Just a tad.

Your data is great.

Olly

[swag72]

I definitely find the first image more pleasing than the second. I find that the second is far too saturated and leaning towards the cartoon. The first image on the other had is subtle and looks far more pleasing to my eyes at least. While it seems that there is always more available to do in processing, sometimes less is more and I think you have demonstrated this perfectly.

The shell is visible in the first image......... what more do you want to see? I'd be interested in hearing from you exactly what you are trying to achieve in the second image and why you feel that you need to do that.

 

[Davey-T]

Beating yourself up again Rodd, you'll get an ulcer at this rate mate 

Your "untreated" image is very nice and I wouldn't do a whole lot more to it and be satisfied.

Trying to enhance stuff that's too faint and can only be caught with more data is pointless, bear in mind also that some "perfect images" that you see may have been captured in ideal conditions, 20hrs captured in my back garden will be worse than 5hrs captured by Olly with the same equipment.

Dave

[gnomus]

Rodd

You've just had advice from two of the best there are.  I'd listen carefully if I were you.....  

[Rodd]

Thanks to all--I do tend to get frustrated.  I see an image shot with similar equipment, processed with PI and I attempt to duplicate it.  I invariably fall short.  

2 hours ago, gnomus said:

Hi Rodd

I think I prefer the first of the two you posted.  The second one is too "full on" for my taste and, in my view,  you are giving up some of the detail in the Ha clouds because all the reds have been turned up to '11'.   How about something in-between (I hope you don't mind me tweaking)?

 

The image I downloaded still had a black point in the 50s and was excessively noisy.  I think with the amount of data you captured you should be able to get a cleaner image if you stretch it less.  I'd still counsel going back to the un-stretched stack and being a little less heavy handed.  The above is just 10 mins or so in Photoshop. 

   

Not at all--looks better.  Don't know what you really mean with the PS talk-I know even less about PS than I do about PI, but I like what you've done.  To be clear--the image that you tweeked (the soft one--the first of the last 2 images) has not been stretched at all--except with the STF (PI speak), which is not a permanent stretch until you lock it in with the histogram-which I did without changing the STF settings at all.  Otherwise, it is as fresh out of the linear state as you can get.  In PI, you STF, then remove gradients with background extraction, neutralize the background, and calibrate the color--all in the linear state.  Then you lock in the STF settings so they are permanent thereby making the image non linear.  But I did not do anything to it after I rendered it non linear.  Maybe I need allot more data--like 20 hours more (10 each of Ha and OIII).  I am looking for clarity, low noise--sharpness, fine detail, like looking through a window into space--what a layman would normally call "resolution" but I don't like to use that word because APs definition of resolution is not what I am talking about.  Anyway.  I appreciate the input and will apply what you, and others , have said.

[Rodd]

1 hour ago, ollypenrice said:

Pixinsight seems to tempt people into roaring mania. No, softly softly. Your first reprocess stops a tad short of what the data has to give. It looks great. The very color saturated ones take it way too far.

I think your gentle image - the one Steve liked - is lovely. I'd give it a slight boost in saturation (slight!!!!!) and bring the greens a tad towards blue. Just a tad.

Your data is great.

Olly

Thanks Olly.  I know what you mean about PI--in my case, I find it hard to stop because I feel as though I don't reach my goal.  I hate the super saturated images too--but I find with my data I can't get sharpness without saturation.  There definitely seems to be a disconnect between my viewpoint and that of others.  To me, I look at my soft, fresh out of linear state image and I see a mediocrity.  I know we've been over this before, but when I click on some of the images on this forum and render them full size, I enjoy surfing through them like I am in a shuttle craft among the stars.  The exaggeration is limited--the pics are that good.  They inspire people like me to dedicate all of their time to trying to duplicate them.  When I render one my images full size, I close it down as quickly as possible!  I don't think its possible for me to explain it--no one seems to understand where I am coming from.  But if you line up 100 people and expose them to my image, and one of the ones I think are really good, my point--what ever it is--will be easily seen and acknowledged by all 100 people.  I am missing something.  My over processing is the symptom.   Maybe I just need more data.  Maybe I need better sky.  

[gnomus]

48 minutes ago, Rodd said:

.... To be clear--the image that you tweeked (the soft one--the first of the last 2 images) has not been stretched at all--except with the STF (PI speak), which is not a permanent stretch until you lock it in with the histogram-which I did without changing the STF settings at all.  Otherwise, it is as fresh out of the linear state as you can get.  In PI, you STF, then remove gradients with background extraction, neutralize the background, and calibrate the color--all in the linear state.  Then you lock in the STF settings so they are permanent thereby making the image non linear......

A-ha.  I wonder if this is what could be amiss.  You are quite correct when you say that you need the STF to see what the other steps do.  However, I find the STF to be far too aggressive to use for the actual stretch.  Knock off the STF (the big red X), then open Histogram Transformation, activate the preview and do a manual stretch.  You can do 2 or 3 stretches and clipping of the black point until you get the effect you like.  Alternatively, don't try a histogram stretch but, after switching off the STF, try the Masked Stretch process instead.  It is better that you take control of the stretch - the STF is too brutal.  If what you have posted is the STF version of the stretch, then there is a very good chance that your data is much, much better than you think it is.    

Have a look at Harry's PixInsight videos if the above is gobbledook.

[gnomus]

If you don't know Harry's videos, they can be found here: http://harrysastroshed.com/pixinsight/pixinsight video html/pixinsighthomenewbie.html  They are all great.  The one I was referring to is the one called "Histogram Transformation".    Harry crams a ton of information into these so watch (and re-watch) closely.

[gnomus]

And just to ram the point home.  Here is one of my efforts with the 'standard' STF stretch applied:

And with a less aggressive stretch (and a few later tweaks):

Your data is considerably better than the top image.  I think that, with a more controlled stretch, you could end up with something quite special.

Steve

[Rodd]

1 hour ago, gnomus said:

And just to ram the point home.  Here is one of my efforts with the 'standard' STF stretch applied:

And with a less aggressive stretch (and a few later tweaks):

Your data is considerably better than the top image.  I think that, with a more controlled stretch, you could end up with something quite special.

Steve

Steve--I do have Warrens videos, but I wil watch Harry's as well.  I did not realize that the STF was overly aggressive.  Makes sense now that you mention it actually.  I know in Nebulosity it takes many stretches to get where you want.  The problem is I found it to be very inconsistent--the STF is so easy and repeatable.  I have never tried doing it manually in PI--other than fooling with the sliders after using STF.  I could never improve on the STF image.  It will take me awhile to learn this approach, but I will reprocess the Crescent.  Many thanks

[Rodd]

4 hours ago, swag72 said:

I definitely find the first image more pleasing than the second. I find that the second is far too saturated and leaning towards the cartoon. The first image on the other had is subtle and looks far more pleasing to my eyes at least. While it seems that there is always more available to do in processing, sometimes less is more and I think you have demonstrated this perfectly.

The shell is visible in the first image......... what more do you want to see? I'd be interested in hearing from you exactly what you are trying to achieve in the second image and why you feel that you need to do that.

 

Sara--thank you for your input.  I'll start by saying that the second result is NOT what I am after.  It is an unfortunate byproduct of me not really knowing what I am doing, so doing it is difficult.  The only tool that I can use that I find improves the image once it is non linear, is curves: saturation and contrast.  I am not very good at getting masks to work, so denoising and sharpening, usually results in a mess.  Same with Pixel Math (except adding Ha to the red channel--I learned that pretty well).  When I look at some of the really nice bicolored Crescent nebulas, the OIII envelope appears clear and transparent, with details of the Ha netting distinct through the blue envelope. There is no fuzz, no blurryness--even when you zoom in.  I have relooked at my "soft" just out of the linear state image, and I agree that it is the best image I posted.  But it is far from awe inspiring.  Maybe it really is my lack of basic processing skills.  The data seems to be good--I have finally managed to chase the hardware gremlins away.  In Pixinsight, there really isn't a whole heck of a lot of discretion at getting where I am with that image.  All I have done with it after calibration and integration is channel combination, dynamic crop, gradient removal, background neutralization and color calibration, then changed it to non-linear--all of which are stock tools used during the linear state in PI.  I have this much down pretty well.  With good data and after proper calibration, these steps take 20 minutes.  I am considering acquiring more data--but part of me says that won't solve my problem.  I think I have enough data to render a nice image--maybe I'm wrong about this too,  though.  More data is always good.  But until I solve the processing problem, all the data in the world might not be the answer.  I'm not sure. 

Anyway, that's the best I can explain why I invariably end up in cartoon land.  Thanks again for the input. 

[swag72]

Processing is the hardest part in my opinion - You have said earlier that you can't seem to be able to replicate other peoples images that you see.... here's the thing...... How about you don't try to replicate them? You can do your own thing, you can develop your own style.

I can't help with anything that you said in the post above.... I'm afraid it's all PI talk and I understand zero of it. What I will say is that processing is a very touchy feely process. I don't think that trying to emulate someone else's picture is the way to learn how to process. You need to look at the data and decide what it actually needs.... sometimes it doesn't need what you may have in your standard tool box and so you need to look again and really decide how you are going to get to an end point.

I guess what I'm saying is that blindly bludgeoning stuff because someone uses this workflow or because their image looks a certain way is not the way to get there in my experience. 

Regarding data, I definitely find processing is easier when there's more of it. There's more room for manoeuvre for sure. I look at peoples images and am often amazed at what they achieve with so little data in my opinion. The project I am currently on will get in excess of 40 hours... Will that make my picture better? It probably won't, but .......... it will make the processing easier for me and the way that I process, and for me that is worth every extra hour that I do.

Being self critical is good and I think that there's too little of that about, but in being self critical one also needs to develop ways in which to get to the final goal.

[ollypenrice]

Two great bits of advice from Sara. 1) Make your own picture. 2) Learn how to look at the data to see what it needs. The second of these is my standard mantra when giving tutorials.

Regarding the stretch, Histogram Transformation, STF and moving the mid point slider left in Photoshop Levels all put a logarrithmic stretch into the data. The auto setting STF simply gives too big a stretch for keeping. It is great for finding gradients, though, because if you can tame them fairly well in a hard stretch then a real world stretch will look fine.

There is another way to stretch, however, and that is not to put in a standard log stretch but to use Curves to shape a stretch of your own. I believe PI now has a Curves option but I use Photoshop. I find the log stretch, probably with the stars masked at first, is fine for colour and very occasionally for luminance. For NB, though, I prefer a far more aggressive stretch, one which gives greater contrasts. This raised a few eyepbrows on here when I posted it but it is from either Adam Block or Ron Wodaski (not sure) and is the technique used by Tom O'Donoghue as welll. Only the first stretch would be as aggressive as this. The second would be gentler.

During tutorials I often do a comparison between this kind of stretch and the 'grey point slider log stretch' and the advantage in terms of structure and contrast is usually convincing.

You can make curves for other purposes, too. Suppose you want to make a starfield background with small stars for a galaxy image: you can just use a curve to bring up the background to the level of the main image and yet keep the stars down. You would then combine this with the galaxy stretch which has large stars.

Personally I find processing in PI to be like trying to mend a watch while wearing boxing gloves. I'm much more at home in Ps. One big difference is that in Ps you have no need to make masks for many tasks. The problem with masks is that it is very hard to make them mask just what you want them to mask. This is important because different parts of an image often need opposite processing intervention. You might want to noise reduce (which means blur) the faint signal while sharpening parts of the bright signal. Layers and the colour selection tool make this so easy in Ps and you can use multiple iterations of the partial eraser to choose exactly where and when to apply the processing modification. You do this while seeing, in real time, what it does.

The PI approach does suit some people but I''m not convinced it's suiting you! (BTW, I do use DBE, SCNR green and sometimes LHE. I do like many aspects of PI but not the global approach.)

Olly

[gnomus]

The other place to find some excellent PixInsight tutorials is here: http://www.lightvortexastronomy.com/tutorials.html  I am with Olly and Sara (who am I to disagree?) - I find PixInsight a very difficult tool to master.  There are folks who get great results with PixInsight alone - check out Barry Wilson's stuff for example.  But with PS I feel I have better control (which means in my case, that I am able to dial down my usual 'bull in a China shop' processing.     

[Rodd]

16 hours ago, ollypenrice said:

Pixinsight seems to tempt people into roaring mania. No, softly softly. Your first reprocess stops a tad short of what the data has to give. It looks great. The very color saturated ones take it way too far.

I think your gentle image - the one Steve liked - is lovely. I'd give it a slight boost in saturation (slight!!!!!) and bring the greens a tad towards blue. Just a tad.

Your data is great.

Olly

 

4 hours ago, ollypenrice said:

Two great bits of advice from Sara. 1) Make your own picture. 2) Learn how to look at the data to see what it needs. The second of these is my standard mantra when giving tutorials.

Regarding the stretch, Histogram Transformation, STF and moving the mid point slider left in Photoshop Levels all put a logarrithmic stretch into the data. The auto setting STF simply gives too big a stretch for keeping. It is great for finding gradients, though, because if you can tame them fairly well in a hard stretch then a real world stretch will look fine.

There is another way to stretch, however, and that is not to put in a standard log stretch but to use Curves to shape a stretch of your own. I believe PI now has a Curves option but I use Photoshop. I find the log stretch, probably with the stars masked at first, is fine for colour and very occasionally for luminance. For NB, though, I prefer a far more aggressive stretch, one which gives greater contrasts. This raised a few eyepbrows on here when I posted it but it is from either Adam Block or Ron Wodaski (not sure) and is the technique used by Tom O'Donoghue as welll. Only the first stretch would be as aggressive as this. The second would be gentler.

During tutorials I often do a comparison between this kind of stretch and the 'grey point slider log stretch' and the advantage in terms of structure and contrast is usually convincing.

You can make curves for other purposes, too. Suppose you want to make a starfield background with small stars for a galaxy image: you can just use a curve to bring up the background to the level of the main image and yet keep the stars down. You would then combine this with the galaxy stretch which has large stars.

Personally I find processing in PI to be like trying to mend a watch while wearing boxing gloves. I'm much more at home in Ps. One big difference is that in Ps you have no need to make masks for many tasks. The problem with masks is that it is very hard to make them mask just what you want them to mask. This is important because different parts of an image often need opposite processing intervention. You might want to noise reduce (which means blur) the faint signal while sharpening parts of the bright signal. Layers and the colour selection tool make this so easy in Ps and you can use multiple iterations of the partial eraser to choose exactly where and when to apply the processing modification. You do this while seeing, in real time, what it does.

The PI approach does suit some people but I''m not convinced it's suiting you! (BTW, I do use DBE, SCNR green and sometimes LHE. I do like many aspects of PI but not the global approach.)

Olly

Thanks Olly--PI does have curves--in fact the only time I use a histogram is when I transfer the STF to the Histogram to make it permanent (non linear).  After that, I only use curves for both contrast and saturation--I never use histogram tool to modify a pic after the initial STF.   But it seems that curves in PS are more diverse than they are in PI (unless I don't know how to use them to their capacity--which is likely I suppose).  If PS did not have a monthly subscription, I would probably try it.  I do have GIMP--downloaded it one day--have never used it.  It is supposed to be Photoshopesque.  Ever hear of it?

[ollypenrice]

I know of Gimp but have never tried it since I use an oldsidh Ps CS3. This subsription business is a nightmare.

Olly