Author Topic: Layers  (Read 22711 times)

Offline budguinn

  • PixInsight Addict
  • ***
  • Posts: 106
Layers
« on: 2010 May 02 08:56:48 »
I hope this isn't considered blasphemous, but I find the layers and mask ability to be one of the most powerful features of PS.

In PS I'm able to perform a very "strong" filter...such as sharpening, noise reduction, saturation, or import a deconvolution.....then hide all. I can then use a brush to only apply to the areas that I'm concerned with.
It completely eliminates the need to build specific masks and the amount of masking can be infinitely adjusted with flow and opacity...then applying a gaussian blur to the mask to eliminate any edges......this is, to me, the ultimate in masking.
I've been fooling with the masking ability of PI and it is more powerful than any of the other programs that I have, but it still doesn't give me the fine control that I'd like to have and can currently do in PSCS2.

Now, is this possible to be implemented in PI?

I know I can workaround this by doing the HDR/deconvolution/transforms...etc in PI and then exporting and importing into PS, but it'd be nice to have a single platform to do it on.

I have no idea what it takes to make a process/interface to do this, but it'd be slick as snot if it could be done.

best regards,

PS, I also like that many of the keyboard shortcuts of PI are the same as in PS...very convenient.

bud

Offline Nocturnal

  • PixInsight Jedi Council Member
  • *******
  • Posts: 2727
    • http://www.carpephoton.com
Re: Layers
« Reply #1 on: 2010 May 02 09:02:10 »
Ah, perhaps you had some traction until you wrote "I can then use a brush to only apply to the areas that I'm concerned with.". *That* is blasphemy :)

Sorry, I don't know enough about PS to suggest alternate workflows that replicate the effect you're after. I always use the luminance mask option in ACDNR so I suppose that's in effect the same thing as a masked noise reduction in PS. Well, process wise, not how the NR works. When I use curves to increase saturation I also use a mask to only do this to the bright areas. So perhaps this all is quite similar?

As your post came in I was viewing one of Harry's excellent videos: http://www.harrysastroshed.com/mask.html If you haven't yet, take a look at it and see if it provides some inspiration.

Glad you're here bud.
Best,

    Sander
---
Edge HD 1100
QHY-8 for imaging, IMG0H mono for guiding, video cameras for occulations
ASI224, QHY5L-IIc
HyperStar3
WO-M110ED+FR-III/TRF-2008
Takahashi EM-400
PIxInsight, DeepSkyStacker, PHD, Nebulosity

Offline budguinn

  • PixInsight Addict
  • ***
  • Posts: 106
Re: Layers
« Reply #2 on: 2010 May 02 09:15:25 »
Hi Sander,

The very first thing I did was watch all of Harry's videos and work with my M106 images at the same time.....super stuff for learning a program.

I have never gotten into the argument (and it is always occurring somewhere on-line) about using the brush method.  I honestly don't see the difference between it and making a mask, such as in PI.  The difference is the control.  The argument is that it is more "art" than true processing.  I think that argument is fallacious....what is the difference in using/making a mask that applies to one little area....then doing it again with a bit more of a HDR and a less discriminating mask....etc........and just doing the full HDR and applying this to only the areas that you want.....completely eliminates all of the trial and error of the "deringing" part......because unless you paint in the stars there is no ringing in the final picture.

As to "art"....that's exactly what it is.....especially when we are trying to emphasize different parts of an image.....at that time (even if it's just bringing out the trap in M42) we have determined what is aesthetically more pleasing to our eye, and how we would like to present this object for the enjoyment of ourselves and others.......and that is art.

It is much like the argument of saturation/color....it is personal preference.

I may have some questions on the debayer routine in the future. 
I have two M25C's and an H9C and I'm just starting to fool with the color conversion.  The SXV and SXVF don't always behave the same when debayered, but I don't see a method for each different style......but, I'm still playing.

best regards,

bud


Offline Carlos Milovic

  • PTeam Member
  • PixInsight Jedi Master
  • ******
  • Posts: 2172
  • Join the dark side... we have cookies
    • http://www.astrophoto.cl
Re: Layers
« Reply #3 on: 2010 May 02 09:16:12 »
Hi Bud

PI's masking system is fully capable of doing that, but perhaps not as "user friendly" or fast.
First of all, you have to remember that here a Mask is any image. And as with any image, you may adjust/transform it as you like (blur, change the brightness, use a constant value [PixelMath allows you to set a constant value for a image]), even when it has assigned as a mask. Mask dependencies are updated each time a new process is thrown, so every time you preview the result, the new mask effects will take place.
You may complain that you cannot see the effects immediately... well, some processes implement a Real Time Preview system, which will allow you to see those effects faster (but with less accuracy, and perhaps you'll need to manually update the status). In result, I think that for masking, or fading purposes, there is absolutly no need for layers. It is just to get familiar with other methods, but the "same" results are obtained.


One final word about brushing. We are against it :) It is like painting, cheating with the image. So, we have to avoid these methods as long as is possible. In most cases, better results are achieved with algorithmic approaches. And, they are much better justified, since there was a "logical" way to segment the image. Manual selections are too arbitrary.
Of course, this is just a "philosophical" point of view about image processing. Painting tools are in our to do list, but with a very little priority level :D
Regards,

Carlos Milovic F.
--------------------------------
PixInsight Project Developer
http://www.pixinsight.com

Offline budguinn

  • PixInsight Addict
  • ***
  • Posts: 106
Re: Layers
« Reply #4 on: 2010 May 02 09:23:40 »
Hi Carlos,

thanks for the very quick reply........and I expected a bit of "hesitancy" when I mentioned painting/brush use......but I hopefully explained my thinking in my reply to Sander.....I think we were typing at the same time.

To me the lack of this isn't a deal breaker at all.....and I've pretty much decided I want the program......

I have never seen such a slick method of revealing the core of M42 without multiple exposures of various lengths.

I look forward to the rest of my playing with this excellent program.

best regards,

bud

Offline Harry page

  • PTeam Member
  • PixInsight Jedi Knight
  • *****
  • Posts: 1458
    • http://www.harrysastroshed.com
Re: Layers
« Reply #5 on: 2010 May 02 09:54:29 »
Hi

I must say in my opion that using a " layers system is not cheating , just using real data where and when you want  >:D saying that I am against painting your image >:(

I mean things like mixing Ha with a RGB image without knackering your stars is hard work :(

So I hope layers is not to far down the list  :footinmouth:

Harry
Harry Page

Offline Niall Saunders

  • PTeam Member
  • PixInsight Jedi Knight
  • *****
  • Posts: 1456
  • We have cookies? Where ?
Re: Layers
« Reply #6 on: 2010 May 02 11:23:25 »
Hi Bud - the way I look at thing is that, if I need/want to use a 'Brush' approach to achieve something that I cannot do in PixInsight, then I happily turn to PS (et al). However, I actually ENJOY trying to find the 'right solution' in PixInsight.

I suppose it is just 'my feeling', but I really do like using ONLY the data 'in' the image to create the 'final' image. And, because of that, I now am actually much happier with my final images even though I know that they could probably be tweaked up (quite) a bit in the likes of PS.

However, that said, it is when the likes of astroimagers such as you 'join' the PI tribe that we will see PI start to come of age. You know what processes you have come to rely on in PS, and you start looking for similar approaches in PI, and don't immediately see them. Then, here on the Forum, you ask questions and discuss alternatives - and perhaps a new processing module is born, a module that benefits all of us. Alternatively, you may realise that there IS an equivalent series of steps in PI itself - which will give you the same result as you were after, but which allowed you, like me, to use the information 'in' the image to help extract the details that you were after.

Enjoy the ride ;D

Cheers,
Cheers,
Niall Saunders
Clinterty Observatories
Aberdeen, UK

Altair Astro GSO 10" f/8 Ritchey Chrétien CF OTA on EQ8 mount with homebrew 3D Balance and Pier
Moonfish ED80 APO & Celestron Omni XLT 120
QHY10 CCD & QHY5L-II Colour
9mm TS-OAG and Meade DSI-IIC

Offline Juan Conejero

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 7111
    • http://pixinsight.com/
Re: Layers
« Reply #7 on: 2010 May 02 15:39:03 »
Hi Bud,

A LayeredComposition tool is on the drawing board and will be released in a future version of PixInsight. I strongly believe that the layers paradigm is interesting and powerful, and definitely it must be available in PixInsight.

However, I disagree with your statement:

Quote
....then hide all. I can then use a brush to only apply to the areas that I'm concerned with. It completely eliminates the need to build specific masks and the amount of masking can be infinitely adjusted with flow and opacity...then applying a gaussian blur to the mask to eliminate any edges......this is, to me, the ultimate in masking.

The idea of drawing a mask by hand with a brush (or equivalent tool) brings up questions such as:

How do you know where you must allow sharpening, deconvolution, etc., and where you don't?

What criteria do you follow to draw a mask with a brush? In other words, if I ask you why a particular pixel has received sharpening, etc., and why others haven't, can you provide a consistent answer?

If you repeat your entire procedure for the same image, drawing your masks at hand with a brush, can you guarantee that you'll achieve exactly the same result?

and so on. I think the way we build and apply masks is perhaps the part of our image processing workflow where our concept of astrophotography becomes more evident, or where it is more clearly shown.

For me, astrophotography is documentary photography. As such, there are limits to what can be done. IMO, the documentary nature of astrophotography is incompatible with arbitrary selections by hand. Selections must be based on properties of the data and must be applied homogeneously to the whole image. This is the difference between a brushed mask and a luminance mask, for example. A luminance mask defines a selection algorithmically, on the basis of a fundamental property inherent in the data: the signal-to-noise ratio is directly proportional to illumination. When we isolate image structures using a multiscale analysis method (a wavelet transform for example), we are also applying a homogeneous criterion based on properties of the data. The same is applicable to classification of image features based on morphological properties, etc. Image processing offers a lot of sophisticated and powerful algorithms suitable for selection of data, so why do you think that drawing a mask by hand is the best way? Because it is more creative?

Algorithmic selections and documentary selection criteria don't limit your creativity. It is just the opposite: with them, you pursue the excellence in your astrophotography by respecting the nature of the data. Applying and implementing algorithmic data selections requires a solid understanding of the data and their behavior with respect to the tools that you are using, which in turn forces you to improve your processing skills and knowledge. In contrast, hand-drawn masks are the easiest path to achieve what you think you have to achieve, without reflection; it is a fast-food solution, a simplification.

For example, there are cases where we perform selections that can be difficult to understand, or even to accept. This M57 image is a good example:

http://pixinsight.com/examples/M57-CAHA/en.html

This image has been criticized because the outer halo of M57 is extremely smooth, while the central region is much more detailed. Naturally, to process the data set that has given raise to this image a lot of masks have been constructed and applied. All masks have been carefully planed and implemented with algorithmic methods and criteria strictly based on properties of the data. The outer halo is thousands of times fainter than the central core of M57. We can only deconvolve the regions of the image where the signal-to-noise ratio is above certain limits, and this is the reason why the external regions have not been deconvolved. Basically, the deconvolution mask is a luminance mask in this case. The 3D representations in the tutorial show this dynamic range problem more clearly. The author's decision has been to show the whole data using high dynamic range compression techniques. I personally applaud that decision, because in this way we can perceive the wonderful nature of this region of the sky and an amazing amount of information that enriches our understanding of the natural objects and phenomena represented in the image, which otherwise would be impossible (i.e., the core would be completely burned out without DR compression), even if this means that we are going to show deconvolved and not deconvolved parts of the image together, with similar brightness, which can be difficult to accept for some people.

Astrophotography is art because it involves interpretation of the data, as the image above demonstrates. Astrophotography is also science and technology. It is documentary photography because its purpose is to show objects pertaining to nature. When I look at an astrophotograph, I need to know that the author has respected --to the extent allowed by the available technology-- the data. I need also to know how the image has been acquired, how it has been processed, and why it has been made and presented as I see it. The final result is, for me, less important than the procedures applied to achieve it.
Juan Conejero
PixInsight Development Team
http://pixinsight.com/

Offline budguinn

  • PixInsight Addict
  • ***
  • Posts: 106
Re: Layers
« Reply #8 on: 2010 May 02 18:38:57 »
Hi Juan,

Thanks for taking the time to respond......

my comments to specific points of yours (in red) are in black....

A LayeredComposition tool is on the drawing board and will be released in a future version of PixInsight. I strongly believe that the layers paradigm is interesting and powerful, and definitely it must be available in PixInsight.

That is very nice.....thanks

However, I disagree with your statement:

Quote
....then hide all. I can then use a brush to only apply to the areas that I'm concerned with. It completely eliminates the need to build specific masks and the amount of masking can be infinitely adjusted with flow and opacity...then applying a gaussian blur to the mask to eliminate any edges......this is, to me, the ultimate in masking.

The idea of drawing a mask by hand with a brush (or equivalent tool) brings up questions such as:

How do you know where you must allow sharpening, deconvolution, etc., and where you don't?
I would say that the same criterion would apply as a person that uses a star mask....deconvolve/HDR transform and see what shows.

What criteria do you follow to draw a mask with a brush? In other words, if I ask you why a particular pixel has received sharpening, etc., and why others haven't, can you provide a consistent answer?
this question could also be asked of the pure mask user.....and is hugely dependent on the type of mask applied

If you repeat your entire procedure for the same image, drawing your masks at hand with a brush, can you guarantee that you'll achieve exactly the same result?
I would say I could do it close, but not necessarily exact....but, using pure masks, would two different processors...totally independent of each other receive "exactly the same result"?..
and so on. I think the way we build and apply masks is perhaps the part of our image processing workflow where our concept of astrophotography becomes more evident, or where it is more clearly shown.

For me, astrophotography is documentary photography. As such, there are limits to what can be done. IMO, the documentary nature of astrophotography is incompatible with arbitrary selections by hand. Selections must be based on properties of the data and must be applied homogeneously to the whole image. This is the difference between a brushed mask and a luminance mask, for example.
But, it would seem to me the whole purpose of a mask is to not apply the process "homogeneously to the whole image"....rather to only that part that is not covered by the mask.....and again, this mask is determined by the input of the human processor predetermined by his desire as to what to enhance and what not to enhance.
A luminance mask defines a selection algorithmically, on the basis of a fundamental property inherent in the data: the signal-to-noise ratio is directly proportional to illumination. When we isolate image structures using a multiscale analysis method (a wavelet transform for example), we are also applying a homogeneous criterion based on properties of the data. The same is applicable to classification of image features based on morphological properties, etc. Image processing offers a lot of sophisticated and powerful algorithms suitable for selection of data, so why do you think that drawing a mask by hand is the best way? Because it is more creative?
An admission......I am ten times better at accenting what I want with a layers mask than using a masking tool, as in PI.  So, is it a creative issue?...I'd say no, it is, to me, waaaay easier.....but, I'll admit, as I use the program more, this may change (hopefully).

Algorithmic selections and documentary selection criteria don't limit your creativity. It is just the opposite: with them, you pursue the excellence in your astrophotography by respecting the nature of the data. Applying and implementing algorithmic data selections requires a solid understanding of the data and their behavior with respect to the tools that you are using, which in turn forces you to improve your processing skills and knowledge. In contrast, hand-drawn masks are the easiest path to achieve what you think you have to achieve, without reflection; it is a fast-food solution, a simplification.

For example, there are cases where we perform selections that can be difficult to understand, or even to accept. This M57 image is a good example:

http://pixinsight.com/examples/M57-CAHA/en.html

This is a knockdown gorgeous image that has been very nicely processed to show the inner details.....just lovely.

This image has been criticized because the outer halo of M57 is extremely smooth, while the central region is much more detailed.
I wouldn't have thought to criticize the image for the halo smoothness.........but, I really do have a pure aesthetic problem with the stars.  So many blooms that were poorly handled...some with vertical blooms that really distract and even some dark horizontal blooms.  These are quite time consuming to fix, but can be done many different ways.....the easiest is to take shorter exposures where the camera didn't bloom and substitute this information for the bloomed information.  My QSI532WSG is a big time bloomer.....but I wanted a camera that was extremely sensitive, and it fits the bill.  So when imaging with this camera I must plan my evening/evenings so that I can work around this eventual defect in the image.
Naturally, to process the data set that has given raise to this image a lot of masks have been constructed and applied. All masks have been carefully planed and implemented with algorithmic methods and criteria strictly based on properties of the data. The outer halo is thousands of times fainter than the central core of M57. We can only deconvolve the regions of the image where the signal-to-noise ratio is above certain limits, and this is the reason why the external regions have not been deconvolved. Basically, the deconvolution mask is a luminance mask in this case. The 3D representations in the tutorial show this dynamic range problem more clearly. The author's decision has been to show the whole data using high dynamic range compression techniques. I personally applaud that decision, because in this way we can perceive the wonderful nature of this region of the sky and an amazing amount of information that enriches our understanding of the natural objects and phenomena represented in the image, which otherwise would be impossible (i.e., the core would be completely burned out without DR compression), even if this means that we are going to show deconvolved and not deconvolved parts of the image together, with similar brightness, which can be difficult to accept for some people.
I don't see the difference between the multiple use of masks and the use of a "Variable Mask" to achieve the exact same end result.....ie  "that we are going to show deconvolved and not deconvolved parts of the image together".  I use a "Variable Mask" and others use a luminance mask....both achieve the same result....using the exact same algorithm for the deconvolution....absolutely no difference in the data, only in what data is revealed.

Astrophotography is art because it involves interpretation of the data, as the image above demonstrates. Astrophotography is also science and technology. It is documentary photography because its purpose is to show objects pertaining to nature. When I look at an astrophotograph, I need to know that the author has respected --to the extent allowed by the available technology-- the data. I need also to know how the image has been acquired, how it has been processed, and why it has been made and presented as I see it. The final result is, for me, less important than the procedures applied to achieve it.

I honestly thought this last statement a bit odd with your previous arguments.......my whole thing is the "final result" ........ie, showing the resultant image as I perceive it and accenting those areas that I desire to highlight.   I also agree with you when you say it is "less important than the procedures applied to achieve it."(This does seem to go completely against your argument for pure luminance style masking).   I don't believe that the use of "variable masks" adds or subtracts data any more than a carefully constructed luminance mask (and I might say, for me, labor intensive and frustrating).  The exact same algorithms that you mentioned are used in both instances of the masks...........however, the "variable mask" method is infinitely more user friendly and IMHO much easier to implement.
I would like to emphasize something, again..........NOTHING is added or subtracted by the use of "Variable masks" that wasn't in the same algorithm that the luminance mask used.  It is the exact same data that underlies both of the techniques.

Now, I hope nobody takes this as any thing more than a simple discussion of ideas/methods in processing.  I don't mean to offend anyone......I usually don't get into these discussions about masks, "real" color, saturation....etc...because someone usually always gets upset and that is not my desire at all.  I'm just trying to learn the program and have some questions about it and  its future directions in addons/capability.

best regards,

bud guinn

Offline Carlos Milovic

  • PTeam Member
  • PixInsight Jedi Master
  • ******
  • Posts: 2172
  • Join the dark side... we have cookies
    • http://www.astrophoto.cl
Re: Layers
« Reply #9 on: 2010 May 02 21:54:01 »
Quote
Now, I hope nobody takes this as any thing more than a simple discussion of ideas/methods in processing.

Of course! We are very glad (and excited) to discuss all of these topics here. This is what the forum is for. You'll find that we are quite open minded ;) :D :D
Regards,

Carlos Milovic F.
--------------------------------
PixInsight Project Developer
http://www.pixinsight.com

Offline RBA

  • PixInsight Guru
  • ****
  • Posts: 511
    • DeepSkyColors
Re: Layers
« Reply #10 on: 2010 May 02 23:49:59 »
Um... Interesting stuff  >:D

For now I just have a message to Bud. Later I'll send a second post...

You said what you think. "They" have said what they think. From your perspective, you're right. From their perspective, they're right.

Well, I think you ALL are right  :P

I have been here and there, and back and forth. I've said many stupid things, and I am in fact doing it right now, again.

Yes, AP is a self-expression exercise.
Yes, AP needs also some self-restraining.
Yes, the limits of that self-restraining are different for every single living thing shooting at stuff late at night.

I once coined this phrase: "There are as many schools of astrophotography as there are astrophotographers". This was in a message I sent to our friendly friend Richard Crisp on some list, who basically replied that nobody would ever quote me on that. So please, feel free to quote me on that ;)

Personally, to say that a localized sharpening throws to the garbage the documentary value of an image that by itself already has a high content of ART and interpretation is just one way to look at things. Not wrong, not right, just an OPINION.

I have learned to believe that painting green a star that originally was blue is WRONG. Most people seem to agree.

But when in a group of 100 people, half of them say that a painted mask is wrong and the other half say it's right, I know I'm dealing with opinions, not facts. Of course, go and make everyone agree...

IMHO, do not try to just limit yourself because others tell you.
Just do what you like.

HOWEVER, I do encourage you to learn about what those others who tell you "nah you shouldn't be doing this" are doing, and their motivation.

You may end up realizing DOWN THE ROAD, that certain things that before you couldn't live without suddenly become BORING, and that what makes you have FUN is something else (I'm not talking about layers BTW). You may end up like Niall, who claims he feels a stronger sense of accomplishment by using only "the data" and he in fact feels he has more fun trying to figure out how to process his image only with "the data". Or maybe not.

But let your own experience to dictate that.

The important thing again, is to keep an eye on WHAT these guys are doing and HOW. I know sometimes it may look ELITIST to some people. It is not. What happens is that some of the things they do are different and cutting edge. To some, that's a lot more FUN and interesting than learning tricks and techniques that have been used for many years. Actually, nothing wrong in learning those "old" techniques! But again, to some, trying to learn/find new and different ways to solve problems is part of the fun.

Juan talked about fast-food. Yeah I've learned to realize that. Sometimes a shortcut seems about right. Problem is, that's not FUN to some people. You said "well I do it because it's waaaay easier!"... I hear you too. But see? Was processing an image "easy" the first time you did it? It probably wasn't. What today may look hard, it won't be hard once you've mastered it (or at least become better at it). But yes, it's hard to try solving a problem with techniques you don't know well yet when you already have a way to do that is "way easier". So have at it sometimes, maybe not all the time... Think you're almost "starting over" and you need to digest tutorials, seek for information that doesn't seem to be anywhere, ask questions here or at other forums... Eventually you'll know how to do things with PI that today you don't even know where to start. And when you have all that knowledge acquired, you will decide what tools you want to use, why, when, and how in order to achieve your goals with any given image. And if a painted mask enters the equation... SO BE IT!!! You'll have your reasons!

Um... Am I lecturing you??? My point is that I believe you've made the right decision by starting to use PI, and you're here asking questions, which is a lot more than many people would do because they're happy living in the "soup of mediocrity" (I didn't coin that term;) ) that sometimes we see in this hobby... And that you don't need to say goodbye to whatever you're doing just because you've been told your images will lose their "documentary value" if you paint a mask (in a minute I'm posting an example of me painting over a mask BTW, I love preaching to the wrong choir!). But do keep all that in mind. Eventually you may find that trying to figure out new solutions to certain problems can be more fun than taking the shortcut. Or maybe not. Or like Niall said: "I actually ENJOY trying to find the 'right solution' in PixInsight."

BTW, my opinion is that in AP, both the methods and the results are equally important. To me.

 O:)


Offline RBA

  • PixInsight Guru
  • ****
  • Posts: 511
    • DeepSkyColors
Re: Layers
« Reply #11 on: 2010 May 03 00:52:59 »
LAYERS
I think Carlos put it nicely when he said "PI's masking system is fully capable of doing that, but perhaps not as "user friendly" or fast."

I agree 100% with that. Also, if I have to choose between a tool that accomplishes X that takes me certain effort and the funometer reads 10, and a different tool  that also accomplishes X but with less effort and for which the funometer also reads 10 or maybe more, and it also doesn't degrade my learning experience, why should I use the tool that's harder to use?

Hopefully my message can be read between the lines :-)

IMHO PS's layering system sucks BTW. And it does because AFAIK, you're stuck with 8-10 "blending modes" and their "opacity" (plus a bunch of effects that we don't use in astro images like glows, bevels, shadows, etc). I doubt PI's layering approach would be so limited, whenever it happens...

MESSING UP WITH MASKS
I've done all sort of crazy things to my images in my "olden" days. Yet, they ended up looking like crap, most of them at least :)
Now I don't nearly mess up with them at all, and surprisingly, they look a lot better.

But... but!!! I'm going to tell you why in my last image for example I PAINTED over a mask and why I don't have any regrets at all...  (doesn't mean you'll agree with me, which BTW is fine with me of course ;) )

Here's a tiny version of a crop of the image (full version here):



Now.. notice the area (top-right) above those two large arches... You can see it's an area with a darker background...
Originally, after I was done with most of my processing, that area was flat dark, in other words, now you can see that despite being darker than the rest, there's some "smoke" in there... (if you stretch the image wildly or see it against a black screen background you'll notice better).

So how did I get all that faint "smoke" to actually be at least somewhat visible? (barely, but visible)...

Quick wrap-up: broke the image into small and large scale structures, applied a histogram adjustment to the large scales image, and added the images back together with a mask... (it's a bit more complex than that, but that's a +/- fair wrap-up). Sounds ok... Now.. Here's how I built the mask:

1.- Extracted luminance
2.- Extracted large scale stuff using ATWT to "blur" it.
3.- Did a histogram adjustment.
4.- Binarized the image to a point where ONLY that area AND a few small black "islands" in other parts of the image were left black - everything else was white.
5.- Blurred the mask with ACDNR
6.- I PAINTED white those smaller black islands.

Juan appeals to the documentary value of an image... Well, here's the thing the way I see it...

After the very reproducible "data based" processes steps 1-5 this is what I ended up with:



You can see the area above the arches nicely unprotected, and you can also see the other "black islands"...

Now, these 1-5 steps may be 100% reproducible and "data based", but just because something is "data based" doesn't make it right. You can do all sort of crazy things with luminance masks!! I personally think I just went overboard with this mask, but hey, it's not painted, so I feel safe ;)

That is not my point though...

Now, let's go back a bit... I am about to apply a histogram stretch with a mask!!! Ok? That's not in my book of supercool things to do. In fact, I assure you that except for protecting the galaxies, no selective stretch has been done on the original image at all during the processing until now....

This means I have already made the decision that I want to brighten some areas in the image while not doing so in others (not only that, I'm going to be stretching only very large scale structures, so I'm even more selective).  But you can ignore that if that already bothers you, because that's not my point either.

Now, by examining in the original image the areas covered by these "black islands" I immediately notice that if I leave those areas unprotected, they are going to suffer. I'll explain... Before this processing, they were darker than their surrounding areas, and show a nice contrast, which I believe is of "documentary value". If I apply this process without protecting those areas as well, I believe I will be degrading the image both, from an artistic and a documentary perspective, because the faint signal in those areas is stronger than the faint signal above the arches, and if left unprotected, the stretch will basically make those areas to blend with their surroundings, as if there was not a decrease in IFN illumination in those areas - which there is. I know... because I tried it.

So I have a decision to make... Either

1) I don't apply this process at all, in which case the faint "smoke" above the arches will remain unseen. I really want those faint details to be a bit visible, and I think that by doing so, the image will have MORE documentary value. It is after all the goal I have with this processing session.

2) I apply it with this mask that has been created "only from the data" but knowing it will degrade areas of the image (the areas under the "black islands") that IMHO have more value, again, from both an artistic and a documentary perspective, if left untouched.

3) I pull the superduper clone stamp and KILL those "black islands" so those areas are not affected from what I'm about to do, and have a crowd of "purists" lynch me in a big way.

Since the goal of this particular processing session was to extract very faint details from where originally we couldn't see any, and if we leave the mask as it is, IMHO the stretch done on those "black island" areas will NOT produce this result, and instead, this processing step will in fact "flatten" those areas  (a case might be made about leaving unprotected the "hole" under the left arch, maybe), it made sense FOR ME - both from an artistic and documentary perspective - to make sure these areas are not affected by this processing session.

So I PAINTED the mask by reducing those "black islands" to pure white and leave unprotected only the area above the arches. And again, IMHO I believe I didn't degrade the documentary value of the image any less than anything done prior to that with a plethora of "data based" masks.

And BTW, I can 100% reproduce this "effect" anytime, though I already knew that 100% replication doesn't guarantee anything.

And of course, I feel right about it because in my very humble opinion, I made the right decision.

And yes, I tried several things prior to pulling the clone stamp... Who knows? It might have been possible to do it without the clone stamp. And I had FUN trying... And I failed. And so I went back to that 3 points decision, and made the decision that I felt was the best. In the end, the improvement is so subtle, I would probably find pointless an argument about my decision.

But if still, anyone would claim that "assuming your image was of documentary value prior to this step, I believe that after doing this it has crossed the threshold of acceptability" my answer would be:

Want to do science with this? Here's my raw data.
Want to give me some feedback? Please do so.
Want to enjoy my work and my interpretation? Look at it.
Want to tell me this image no longer is acceptable AP? Remind me not to discuss this topic with you again :P

And they lived happily ever after, the end  8)


EDIT: BTW I forgot to mention... In the end, I toned down the effect of this modification A LOT, so in my final published image, the "stuff" above those arches is much less noticeable than in the screen shot above. Funny things the mind does sometimes :-)

EDIT 2: And the screenshots above are focused only on that area. To see my final published image click here...

EDIT 3: And the screenshot of the mask (which covers the full image) does NOT match the first screen shot - which is a crop. Which sucks because then you cannot see well the correlation between the "black islands" and the area they cover in the real image, but hopefully you'll figure it out by using the link I gave in EDIT 2.  SORRY!!!


« Last Edit: 2010 May 03 01:12:30 by RBA »

Offline georg.viehoever

  • PTeam Member
  • PixInsight Jedi Master
  • ******
  • Posts: 2132
Re: Layers
« Reply #12 on: 2010 May 03 01:16:10 »
I think it is as usual: It is not the tools that mess up an image, it is the user who does it. It is perfectly possible to generate "painted" images using algorithmic procedures only. As a contrived example: Imagine a process that is capable of selecting the pixels that belong to a spiral galaxy to be used as a mask. Such an algorithm is perfectly conceivable, given the special properties of spirals. Now, what is the difference to a carefully "hand-painted" mask? Just because it has been generated manually does not mean it is a bad thing compared to the automatically generated mask.

Just my 2 cents,
Georg

Georg (6 inch Newton, unmodified Canon EOS40D+80D, unguided EQ5 mount)

Offline Harry page

  • PTeam Member
  • PixInsight Jedi Knight
  • *****
  • Posts: 1458
    • http://www.harrysastroshed.com
Re: Layers
« Reply #13 on: 2010 May 03 01:30:06 »
Hi People


great discussion  O:) and as long as its objective and no finger pointing takes place no one is going to get upset and its good  to hear from the skilled   :D
 
 and you gents can type better and faster than me ( Must try and use more than one finger )

Just to say to me Painting a mask is Just fine , its not painting the image IMO  >:D

Bud , I am sure once you have got passed my basic introduction to pixinsight you will beat it into submission  ::)

Harry
Harry Page

Offline vicent_peris

  • PTeam Member
  • PixInsight Padawan
  • ****
  • Posts: 988
    • http://www.astrofoto.es/
Re: Layers
« Reply #14 on: 2010 May 03 02:03:29 »
Hi,

to me this is a simple matter. A mask must follow two (two, not one of them) principles:

- It must built with purely documentary arguments (these arguments may include aesthetic reasons, but the first is essential).

- It must cause an "evenly uneven" processing of the data (this is what Juan means): we must have the same criterion at time of selecting a set of data inside the whole image.

These two principles have nothing to deal with tool, but with imagers, and with the question you must have always in mind: "Why am I doing this?"

Painting masks has an additional problem. You paint the mask because it's quick and easy, and because you can skip some technical details when selecting a data set. This will come against you if, in the future, you have a major problem to solve.

Don't worry about using a determined tool. Always worry about how and what are you doing.


Regards,
Vicent.