Image processing theory or artistic skills

[souls33k3r]

Hi All,

Thought it's ask this.

When processing an image, how much does the theory and how much artistic skills play a role in image processing?

What i have seen mostly is people are really good with theory but they might dwell so much in to the theory of everything that the results are not as eye pleasing but then we have people who don't really care as much about theory but just plain and simple try a different slider, a different number unless it's an eye pleasing result.

I understand that good data is everything but then how do you quantify good data without going in to the nitty gritty of the details? I also understand you need to know the tool you're using but does it really matter the mathematics behind it?

Thanks

Regards,

 

[PhotoGav]

Personally I am trying to create a pleasing image that remains faithful to the data captured. I don't want to stray unrecognisably from what the target 'really' looks like. Of course any image is an interpretation of the data, with tweaks to colour, contrast, sharpness, brightness and dynamic range, etc., but, I think the image is most interesting when it shows us what is there that we simply couldn't see with our own eyeballs. So, understanding what the software is up to is important, but understanding how the software does it, I believe, is unneccessary and also probably beyond my realm of understanding! However, doctoring the image to the point of it being a work of art rather than a beautiful representation of the data captured is, in my opinion, pointless.

[souls33k3r]

I would most definitely put myself in the same category as you Gav. In short, as long as it looks pretty but not over cooked. 

But how do you quantify data being "good" for that?

[upahill]

I lean more towards the 'press it and see' method, primarily because im still learning the theory. There needs to be some creativeness in the image - otherwise are we all just striving to recreate the same perfect image?

[michaelmorris]

15 minutes ago, PhotoGav said:

However, doctoring the image to the point of it being a work of art rather than a beautiful representation of the data captured is, in my opinion, pointless.

Personally, I too don't like images that are manipulated to the point of not looking 'naturalistic'.  However, (playing devil's advocate at bit) I think could be a place for 'doctoring the image to the point of it being a work of art rather than a beautiful representation of the data captured. '  JMW Turner's numerous depictions of sunsets, Van Gogh's 'Starry night' and countless works by the French impressionists are all lousy 'true representations' of the scene depicted, but are generally hailed as beautiful works of art.  Are these pointless? 

[souls33k3r]

1 minute ago, michaelmorris said:

JMW Turner's numerous depictions of sunsets, Van Gogh's 'Starry night' and countless works by the French impressionists are all lousy 'true representations' of the scene depicted, but are generally hailed as beautiful works of art.  Are these pointless? 

I never understood art if i'm being honest especially those you mentioned above.

[PhotoGav]

16 minutes ago, souls33k3r said:

I would most definitely put myself in the same category as you Gav. In short, as long as it looks pretty but not over cooked. 

But how do you quantify data being "good" for that?

The data needs to be deep enough to bring out the faint details, exposed correctly not to burn out the bright details, composed well to create an interesting image, tracked perfectly to avoid any trailing or blur, set up perfectly to be sharp across the whole field of view, have enough subs to combat noise, all be correctly calibrated to deal with all the other sorts of noise and artefacts and stacked perfectly to make the best possible use of all of the above! And then you start tweaking it to make it look pretty...!

20 minutes ago, upahill said:

I lean more towards the 'press it and see' method, primarily because im still learning the theory. There needs to be some creativeness in the image - otherwise are we all just striving to recreate the same perfect image? 

Absolutely, I agree. However, I think we are all striving to 'create' the same perfect image. Unfortunately nobody has quite managed it yet, so 'recreate' doesn't apply yet!!! When it comes to it, the 'same perfect image' is never quite the same is it - different set ups, different framing, different subs, different processing to bring out different aspects of the given target. In the words of Groove Armada 'if everybody looked the same, we'd get tired looking at each other' - I think that sort of applies here!

2 minutes ago, michaelmorris said:

JMW Turner's numerous depictions of sunsets, Van Gogh's 'Starry night' and countless works by the French impressionists are all lousy 'true representations' of the scene depicted, but are generally hailed as beautiful works of art.  Are these pointless?

Absolutely not. The artists in question set out to create art works, not astrophotographs that show what is out there but not visible with the human eye. There is most definitely a place for using telescopes and cameras to create beautiful artworks, it's just that that is not what I am using the kit to try and do.

[michaelmorris]

Getting back to the issue in the original posting, I think that understanding the mathematics behind the data manipulation algorithms is not a prerequisite to creating great astro images.  However, I think understanding at least the basics how a tool or technique works should do nothing but help any engineer, craftsman or artist have a greater appreciation of the capabilities and limitations of that technique or tool.

I have for a while been intrigued that there are many books and webpages devoted to the use of tools and techniques in astrophotography, but precious few  deal with the subject of 'How to make a beautiful image'.  There is little on subjects such as target selection, image composition, dynamic range, colour and/or texture.  For instances, Rob Gendler's excellent 'Lessons from the Masters' book contains just one chapter on image composition.  Most astrophotography books and webpages don't seem to mention these subjects at all.

[souls33k3r]

4 minutes ago, PhotoGav said:

The data needs to be deep enough to bring out the faint details, exposed correctly not to burn out the bright details, composed well to create an interesting image, tracked perfectly to avoid any trailing or blur, set up perfectly to be sharp across the whole field of view, have enough subs to combat noise, all be correctly calibrated to deal with all the other sorts of noise and artefacts and stacked perfectly to make the best possible use of all of the above! And then you start tweaking it to make it look pretty...!

Absolutely, I agree. However, I think we are all striving to 'create' the same perfect image. Unfortunately nobody has quite managed it yet, so 'recreate' doesn't apply yet!!! When it comes to it, the 'same perfect image' is never quite the same is it - different set ups, different framing, different subs, different processing to bring out different aspects of the given target. In the words of Groove Armada 'if everybody looked the same, we'd get tired looking at each other' - I think that sort of applies here!

Absolutely not. The artists in question set out to create art works, not astrophotographs that show what is out there but not visible with the human eye. There is most definitely a place for using telescopes and cameras to create beautiful artworks, it's just that that is not what I am using the kit to try and do.

Very well said at all fronts Gav. One question though, you mentioned tracked perfectly to avoid any trailing. How much trailing can you really get away with and how would it affect the image in the end? Do you think a single sub elongation is negligible if the stacked image turns out with round stars?

 

3 minutes ago, michaelmorris said:

Getting back to the issue in the original posting, I think that understanding the mathematics behind the data manipulation algorithms is not a prerequisite to creating great astro images.  However, I think understanding at least the basics how a tool or technique works should do nothing but help any engineer, craftsman or artist have a greater appreciation of the capabilities and limitations of that technique or tool.

I agree with the basics but does it help to navigate in to the numbers games to find out if the result you're trying to achieve is not yet possible because the numbers aren't good enough or you carry on regardless as long as Mak 1 eye ball says it's good?

[PhotoGav]

6 minutes ago, souls33k3r said:

One question though, you mentioned tracked perfectly to avoid any trailing. How much trailing can you really get away with and how would it affect the image in the end? Do you think a single sub elongation is negligible if the stacked image turns out with round stars?

One dodgy sub is always going to bring the whole stack down, so ideally you would not use that sub. However, using one more sub can have a positive impact on reducing noise, which might ultimately be more beneficial. As with all of this, it is a question of experimenting, trying with and without and seeing what the difference is. In the end, I reckon an image that is just a bit below perfect is way better than no image at all because it wasn't perfect!

[souls33k3r]

2 minutes ago, PhotoGav said:

One dodgy sub is always going to bring the whole stack down, so ideally you would not use that sub. However, using one more sub can have a positive impact on reducing noise, which might ultimately be more beneficial. As with all of this, it is a question of experimenting, trying with and without and seeing what the difference is. In the end, I reckon an image that is just a bit below perfect is way better than no image at all because it wasn't perfect!

Very well put mate.

[Jonk]

You can apply the same question to daytime photography.

Very rarely are published photos a true representation of reality, although for sports media I think that they have to literally show as captured, so no tweaks to contrast etc.

For me, all I can go on are previous astro images, so if I get enough 'good' data on a target and follow the ideas of others, I can create something that looks similar.

Adding information (painting) is certainly a no-no, but how may times do people do spot correction for example?

An astro image in my opinion is never finished - in fact, I don't think in the short time I've been growing into astrophotography I have ever 'finshed' an image!

I do have a few images on my PC background that are clearly 'abstract art' rather than real captured data - are they any less impressive...?

 

[vlaiv]

I'm in the "numbers" camp

For me astrophotography has definite roots in astronomical data collection, and I tend to look at astrophotograpy as visualization of relevant data rather than art in it self. So data comes first for me, and with such approach, I put higher value on understanding what data manipulation I'm doing in order to create visual representation of data, rather than artistic image. Great thing about such data manipulation is that we can end up with images that have artistic feel and are pleasing to the eye, but I believe first and foremost it should be about conveying information about subject (structure, relative position to other things, dynamic range of emitted / reflected light, sense of scale / size, etc). This is probably artistic aspect of data representation, as there are no rules how one should best "interpret" data to convey such a message.

To put it simply - I would rather do "scientific" method of color balancing an image (if doing RGB, like photometric matching of star colors to obtain valid colorimetric data, then applying various established transforms with respect to presentation medium - like PC monitor, etc ...) than trying to adjust colors to make a good color composition or impair certain "feel" to the image.

On the other hand when doing NB, I would rather use custom color palette (rather than standard SHO, or any direct mapping type) if it would help interpret "depth" or other characteristics of recorded object (gas composition, structure, whatever). It is after all "false color" image, and I expect people to recognize that hue carries different information about subject than mundane one.

[souls33k3r]

1 hour ago, vlaiv said:

I'm in the "numbers" camp

For me astrophotography has definite roots in astronomical data collection, and I tend to look at astrophotograpy as visualization of relevant data rather than art in it self. So data comes first for me, and with such approach, I put higher value on understanding what data manipulation I'm doing in order to create visual representation of data, rather than artistic image. Great thing about such data manipulation is that we can end up with images that have artistic feel and are pleasing to the eye, but I believe first and foremost it should be about conveying information about subject (structure, relative position to other things, dynamic range of emitted / reflected light, sense of scale / size, etc). This is probably artistic aspect of data representation, as there are no rules how one should best "interpret" data to convey such a message.

To put it simply - I would rather do "scientific" method of color balancing an image (if doing RGB, like photometric matching of star colors to obtain valid colorimetric data, then applying various established transforms with respect to presentation medium - like PC monitor, etc ...) than trying to adjust colors to make a good color composition or impair certain "feel" to the image.

On the other hand when doing NB, I would rather use custom color palette (rather than standard SHO, or any direct mapping type) if it would help interpret "depth" or other characteristics of recorded object (gas composition, structure, whatever). It is after all "false color" image, and I expect people to recognize that hue carries different information about subject than mundane one.

If you're a student of the game then yes numbers is where the real science lives but most of us are here because we choose this as a hobby to image the night skies and paint it with pretty colours be it false. So we like to play around with our data like little kids on a beach with buckets and shovels building a sand castles. 

Most of the APOD's (if not all) that i've seen are based on someone's artistic approach to the data. I don't know, maybe the numbers game does indeed provide the necessary excitement to many but i'm in the other camp.  

[Jonk]

I agree with that, I couldn't really care less why the colours are there, I'm interested in the fact that they are there and how big / far away they are.

[vlaiv]

39 minutes ago, souls33k3r said:

If you're a student of the game then yes numbers is where the real science lives but most of us are here because we choose this as a hobby to image the night skies and paint it with pretty colours be it false. So we like to play around with our data like little kids on a beach with buckets and shovels building a sand castles. 

Most of the APOD's (if not all) that i've seen are based on someone's artistic approach to the data. I don't know, maybe the numbers game does indeed provide the necessary excitement to many but i'm in the other camp.  

I understand that many people might not be interested in "numbers" aspects of this hobby, but don't dismiss them as being too "sciency", we are still talking hobby level here, and since we all choose hobby that is inquisitive in nature and very much tied to its scientific background, so there is plenty of aspects to choose from, why not give "numbers game" a go? You may find that you don't like it, or your interests may shift with time, but don't dismiss aspects of hobby based on the fact that they sound too much "scientific" (and hence no fun ) - after all you might find additional satisfying dimension to this hobby.

[souls33k3r]

2 minutes ago, vlaiv said:

I understand that many people might not be interested in "numbers" aspects of this hobby, but don't dismiss them as being too "sciency", we are still talking hobby level here, and since we all choose hobby that is inquisitive in nature and very much tied to its scientific background, so there is plenty of aspects to choose from, why not give "numbers game" a go? You may find that you don't like it, or your interests may shift with time, but don't dismiss aspects of hobby based on the fact that they sound too much "scientific" (and hence no fun ) - after all you might find additional satisfying dimension to this hobby.

Oh don't get me wrong mate, there's nothing wrong with the science part of it. It's like a friend of mine, all what he does is images asteroids and he's at the very top of the asteroid hunting food chain. His images are basically black dot on a white background but what he goes is amazing. His data contributes towards the science of it. Unfortunately, it's not my thing but it's his and i'm happy for him

[carastro]

I am in the camp of "try this and see if it works" - having previously learned many processing skills from some very knowledgeable teachers.  I don't analyse anything mathematically, probably because I can't.  

I process images until I am happy with how they look and this is why Photoshop works for me as I can see what I am doing. 

I do not add anything that wasn't already there, though I might take away something such as a stray hot pixel or satellite trail.  I might increase colour saturation, and stretch the data and may do this in several layers taking out the best from each.  At the end of the day I hope to get a pleasing image, both for me and my peers.

Carole 

[Xiga]

Interesting topic this. Thanks @souls33k3r! I'll give my 2 cents. Well, 2p actually, as i'm technically in the UK! 

It's funny, everything about me, my personality, my job, my interests...i'm a numbers and science guy. But when it comes to AP, i'll always trust my eyes over my brain. I don't like the thought of changing a setting with some random numerical value, just because i read it in an article somewhere, i'd always rather test things out in real-time with my own eyes and decide for myself what i think looks best. And for this, nothing will ever beat Photoshop. This is why i will always have PS in my post-processing workflow. Most likely, as the years pass by, it might play an ever smaller role, but it will always be in there somewhere for sure (i keep casting sideways looks at PixInsight, but until it works out how to better explain to me, in a sensible manner, all of it's myriad settings then i'll continue to pass.....for now!)

As for the Arty side of the debate, i'm a little torn because it's a bit of both for me, but if i had to choose one or the other, then i'd say i'm more on the Art side of things (but with the caveat that the final image mustn't deviate significantly from what it should look like). 

Imagine if the post-processing element of the hobby was dealt with in exactly the same way as we do the pre-processing element. i.e a mostly set-in-stone scientific process that is numbers-based, with little to no deviation. Then we'd all end up looking at virtually the same images. How dull would that be?! 

Actually i've just had a worrying thought. With the advent of AI, i wonder how long it will take someone to develop a program that will use AI to fully process an image. When you think about it, given what is already possible right now, how hard would it be to write an algorithm that learned from watching an expert PixInsight user process a series of images that covered most eventualities. Scary thought! Would you be happy or sad to not have to bother processing your images any more? (i'd definitely be sad!). And to take it even further, if you are one of the lucky ones who has a fully automated setup, then technically speaking you could be completely removed from the whole process altogether! Now that's what i call a hobby! Lol. 

Sorry, i deviated a bit towards the end there. Just having some fun ?

[carastro]

Quote

 if you are one of the lucky ones who has a fully automated setup, then technically speaking you could be completely removed from the whole process altogether!

Good point.  We are in this hobby for enjoyment, so although it is great to make life easier for ourselves in certain aspects, I think to take away all the personal in-put and taste and feeling of achievement,  would as you say leave you with no involvement in the hobby and then "what is the point"?  Lol 

Carole 

[souls33k3r]

9 hours ago, Xiga said:

Interesting topic this. Thanks @souls33k3r! I'll give my 2 cents. Well, 2p actually, as i'm technically in the UK! 

It's funny, everything about me, my personality, my job, my interests...i'm a numbers and science guy. But when it comes to AP, i'll always trust my eyes over my brain. I don't like the thought of changing a setting with some random numerical value, just because i read it in an article somewhere, i'd always rather test things out in real-time with my own eyes and decide for myself what i think looks best. And for this, nothing will ever beat Photoshop. This is why i will always have PS in my post-processing workflow. Most likely, as the years pass by, it might play an ever smaller role, but it will always be in there somewhere for sure (i keep casting sideways looks at PixInsight, but until it works out how to better explain to me, in a sensible manner, all of it's myriad settings then i'll continue to pass.....for now!)

As for the Arty side of the debate, i'm a little torn because it's a bit of both for me, but if i had to choose one or the other, then i'd say i'm more on the Art side of things (but with the caveat that the final image mustn't deviate significantly from what it should look like). 

Imagine if the post-processing element of the hobby was dealt with in exactly the same way as we do the pre-processing element. i.e a mostly set-in-stone scientific process that is numbers-based, with little to no deviation. Then we'd all end up looking at virtually the same images. How dull would that be?! 

Actually i've just had a worrying thought. With the advent of AI, i wonder how long it will take someone to develop a program that will use AI to fully process an image. When you think about it, given what is already possible right now, how hard would it be to write an algorithm that learned from watching an expert PixInsight user process a series of images that covered most eventualities. Scary thought! Would you be happy or sad to not have to bother processing your images any more? (i'd definitely be sad!). And to take it even further, if you are one of the lucky ones who has a fully automated setup, then technically speaking you could be completely removed from the whole process altogether! Now that's what i call a hobby! Lol. 

Sorry, i deviated a bit towards the end there. Just having some fun ?

I completely agree with you on that @Xiga but then that's my personal opinion and i totally get it why people who've used PS for such a long time are the best advocates of the software but i've started my AP journey through PI and can understand that most of the stuff that goes on in the back ground has a mathematical value against it making it a powerful tool but i guess it's just too powerful to lead you in to the ibis (i might get stoned to death for saying that) but i'd still like to learn more about the software, what setting makes what ticks and bring out the best from the end result but still keep it very artistic and close to the hubble pallet in my mind and not go overboard.

The AI business is indeed very scary, i personally would be scared of that because it's taking most of the fun away from us. Don't get me wrong, we love some level of automation but not enough for the software to open the roof for you, choose the target, no guide scope in the mix, take light and calibration frames and close the roof in the end. To me, where's the fun in that? and then to top it off with AI process? eek!