Improving the Camera: The Image

[Sabovic Adis]

Ah, drugs !

P.M. me! :rofl:

No, really, if You don't believe what's written You'll have to check it for Yourself, by Yourself. For example, if 35 MP sensor really is a 35 MP sensor. Counting pixel per second, eight hours per day, and that for twenty days, gets You at about 1 MP. Good luck! :lurk5:

 

[Sabovic Adis]

As such please change your description at the beginning of this thread and remove the link to the product. Either it does what you claim it can or it's false advertising.

And this could be calumny...

 

[Hrvoje Simic]

As such please change your description at the beginning of this thread and remove the link to the product. Either it does what you claim it can or it's false advertising.

Yo Aharon, I see the post count has risen recently due to obvious care for fellow men and wish to improve my understanding of the matter. Very noble.

Tell you what...if you give me your email and/or IP address I will gladly block them from the ability of acquiring anything ever from my project, thus helping you to avoid the risk of being tricked, mislead and parted from your money due to someone's selfish desires.
I can help Mike as well.

 

[Jaime Vallés]

This all looks fantastic, Hrvoje. Any plans on making Primers for Canon cameras? My 1DX Mark II could use some love.

 

[Joel Arvidsson]

Dont give him any ideas before he finish the helium primes

 

[Hrvoje Simic]

This all looks fantastic, Hrvoje. Any plans on making Primers for Canon cameras? My 1DX Mark II could use some love.

Canon love soon...for now C100 and C300.

Dont give him any ideas before he finish the helium primes

No worries...

 

[Aharon Rothschild]

Still waiting on Hvroje to acknowlrdge that LUTs are destructive. Literally the entire advantage of shooting RAW is the ability to make non destructive conversion and gamma/color correction before primary grade.

 

[Hrvoje Simic]

Still waiting on Hvroje to acknowlrdge that LUTs are destructive.

Still waiting for you to roll over and do a backwards flip because I said so and find it useful for some reason but I'll save that condition, the realization of which I am of course entitled to, once you actually develop something potentially useful to others and share it on these boards, while questioning your motives and knowledge because it has some price attached. Also, I propose checking the correct name spelling when you address people in a discussion.

No one said or suggested the opposite so pointing the obvious has zero sense.
Everything you do to a digital image is destructive from the moment you touch it, every computing process has micro computing errors, which is why 32 bit float precision is used to reduce the effect of those errors to negligible levels.
It depends on tools and how they are used if those destructive outcomes will become visible and problematic or not. You can brake the image with every image manipulation tool you use. It doesn't matter if it's raw or rgb, log or linear, those factors only determine the range, aspects and types of consequences of image manipulation.

Suggesting that all lookup tables should be avoided due to "destructiveness" is just generalization and FUD.
It greatly depends on how the image values are mapped in a lookup table and one of the reasons this process takes so long is testing these tools in various scenarios and doing multiple optimizations for real world usage.

Literally the entire advantage of shooting RAW is the ability to make non destructive conversion and gamma/color correction before primary grade.

That has absolutely nothing to do with Primers and there is zero conflict with that.
You continue to base your rhetoric as if you are talking to a complete ignorant.

Had you actually read the thread and invested some effort in understanding it, instead into a continuous battle with your own strawmen, that would already be clear, this back and forth avoided and time and energy of two of us and people having to read this saved for something else.

You do the appropriate adjustments in camera raw settings, and appropriate color grading adjustments in log before the Primer and appropriate after the Primer.
That way you keep the advantage of raw, and of log based grading while (synthesis) you harvest the precise and particular signal optimization possible with a per-camera designed image transformation, wrapped in and defined by 262,144 signal transformation values.

And yes, someone actually cares and invests own time and effort so those transformations don't mess up your image but improve it in many neutral focused aspects so you could spend more resources on creative work, instead of using a shovel for micro surgery by continuously attempting to fix things which have already been fixed. If you really really want to do that, I won't try to stop you or butt in your preferred m.o. forcing you to do otherwise.
I can only offer a route.

 

[Sabovic Adis]

Still waiting on Hvroje to acknowlrdge that LUTs are destructive.

No. They're just digital.

Literally the entire advantage of shooting RAW is the ability to make non destructive conversion and gamma/color correction before primary grade.

Shoouting RAW ain't got nothing to do with LUTs.

 

[Hrvoje Simic]

Dragon.

Luigi.

No grading done.



ISO 800
WB 5600K


Left to right:

RG3/DC2 ---- Red IPP2 Log3G10/RWG LUT ---- Primer NTR on Log3G10/RWG + touch of contrast.

Good luck with grading to get to skin palette in the third example.






ISO 800
WB 4200K


Red IPP2 Log3G10/RWG LUT ---- Primer NTR on Log3G10/RWG + touch of contrast.

ISO 800

WB 5600K (missed, more accurate is 5900K)


Red IPP2 Log3G10/RWG LUT

Primer NTR on Log3G10/RWG + touch of contrast.

 

[Wayne Morellini]

No. They're just digital.



Shoouting RAW ain't got nothing to do with LUTs.

I think you misunderstand Aharon. It depends on how you do it. Luts applied and the result baked in, when pushed around a second time to set a look, you can loose some original information. But before you bake it, a lut applied at display time doesn't affect the original data. So, each time you bake it you can loose some room to move. It matters because of constrictions in the workflow. So the reality has been in setting as much look during shooting allows more flexibility around that starting point in post, as it's generally better to do these things in camera and be left with a file with a camera file full range around that look. But, who knows what you eventually want. With something like helium, you should be starting out better anyway. Whatever Hrvoje is doing it should be adjustable, as a starting point before being baked in.

 

[Sabovic Adis]

I think you misunderstand Aharon.

Absolutely, Wayne, and I have a very good reason to do so.

It depends on how you do it. Luts applied and the result baked in, when pushed around a second time to set a look, you can loose some original information. But before you bake it, a lut applied at display time doesn't affect the original data. So, each time you bake it you can loose some room to move. It matters because of constrictions in the workflow. So the reality has been in setting as much look during shooting allows more flexibility around that starting point in post, as it's generally better to do these things in camera and be left with a file with a camera file full range around that look. But, who knows what you eventually want. With something like helium, you should be starting out better anyway. Whatever Hrvoje is doing it should be adjustable, as a starting point before being baked in.

I agree, more or less. :biggrin:

 

[Aharon Rothschild]

This is so interesting to me as I use LUTs all day every day on something like 30 features and have a strong understanding on what they can and cannot do.

I've said all along that the usefullness of this LUT depends on how destructive it is, and that is a fair and balanced assessment. The back and forth here is some sort of cognitive dissonance because the premise of a primer has unwanted consequences that I'll explain.

Gamma and color correction as part of the debayer are the only reason we put up with files that need a red rocket or insane CPU power to play in realtime. Non destructive correction means changing color temperature 100 times with no image degradation. The color science going in to the color and gamma tranforms that are part of the SDK and the debayer are iterations that have evolved since the red one and reflect a effort by Graeme and others to give the user a "optimised" signal etc. Hrvoje sounds like he's describing IPP2 http://www.postmagazine.com/Press-C...introduces-image-processing-enhancements.aspx

It is far pereferable to use red color science in a non destructive way for this purpose. In my opinion LUTs have a very important place as artistic or look LUTs as a easy way to access what is essentially a saved correction. There is no issue with shaping color with LUT's in a destructive way as there is no issue with shaping color with lift gamm gain. On the other hand putting a LUT out like this and claiming that it is a camera "improver" is disingenuous at best and reveals a basic misunderstanding about the process. Once the image has been debayered there is no improving, only changing and every change this "primer" makes has a consequence. Just on the face of it the before and after corrections we see saturated blue skies in some images and overly magenta skin in others, there's a connection there...

Explained another way LUTs are useful as part of a overall correcriron pushing towards a particular look. Grading on top of a corrector LUT (coming from a image that started with great color science) is a very bad idea as the LUT has limited color with it's correction and changed it in a global way non specific to the shot at hand. Basically you are trading red color science for Hvroje color science. Hvroje says that the change has eliminated image color contamination making for easier correction downstream and optimised the image, but if so it's a open question as to why those deviations where present in the debayered signal. I would guess that someone has already done this work and arrived at a more sophisticated solution namely the existing Red color science.

To sum up: All luts are destructive. They are best used as part of a look and have little utility at the base correction stage beyond correcting for gross color casts or matching cameras or displays. AS they correct they change the image in a destructive "locked in" way and as such are not well suited to correction which is a flexible process that adresses each shot individually.

 

[Sabovic Adis]

Scholastic nonsense!
How destructive are they? Quantify destructiveness and state measuring unit of it!

 

[Aharon Rothschild]

Scholastic nonsense!
How destructive are they? Quantify destructiveness and state measuring unit of it!

Its a term used to differentiate between types of correction. In this instance the amount of change made to the image in a permanent "locked in way"

 

[Sabovic Adis]

Its a term used to differentiate between types of correction. In this instance the amount of change made to the image in a permanent "locked in way"

Yes, amount. How much?

 

[Aharon Rothschild]

That is like asking for the temperature of the word "hot"

 

[Aharon Rothschild]

Its a process not a amount. Perhaps google the subject?

 

[Hrvoje Simic]

I am thankful for the test of patience.


...

This is so interesting to me as I use LUTs all day every day on something like 30 features and have a strong understanding on what they can and cannot do.

Always interesting to see self-valorization and the motivation behind it.

Okay...so...your "strong understanding" of lookup tables is based only on their application and of the ones you came across ?


= >

Barbara
Celaren
Darrii
Ferio
...

https://en.m.wikipedia.org/wiki/Syllogism

I've said all along that the usefullness of this LUT depends on how destructive it is, and that is a fair and balanced assessment.

Good thing someone invested own time and energy to improve things and minimize the destructiveness.

You haven't been saying this all along.

Had you approached it with this rhetoric there would be no reason for neverending babble taking my time which could have been used for improving things which are not constantly dropping out of someone's head. You are systematically trolling the thread from the moment you joined the discussion, by using every possible straw you may come up with to devaluate the project, having zero experience with what it's about. Everything here is of course questionable and you know better. Good luck with that reasoning and Karma it unavoidably brings.

That "fair and balanced assessment" is what you write now reformulated as damage control with an assumption the former will pass unnoticed and you actually benevolent and friendly. You haven't been neither friendly, nor considerate but 100% dismissive.

The back and forth here is some sort of cognitive dissonance because the premise of a primer has unwanted consequences that I'll explain.

a) Cognitive dissonance is a creation of your own mind and has nothing to do with me. I've invested effort in helping you understand, even warned at the very start of the thread about the required properties to assimilate this, knowing what departure from conventional it takes, having had to rewire my brain 100 times in the last two years to pull this off. I even pointed to you mental processes leading to these misconceptions and errors in reasoning. You just don't seem to want to listen and consider and just continue giving lectures of LUTs on and on, dismissing everything what was written and assuming all those lectures are really necessary.
All this time I am trying to help you to avoid this dissonance but success of that endeavour requires a two-way effort.

b) Names are typically written with a capital letter.

c) You didn't understand the premise and you are guessing and guessing wrong and have zero experience with the tool discussed so any attempts of explaining this matter by addressing other things are yet another logical fallacy.

d) Tools are already implemented in a few serious projects around the globe with feedback from people with actual experience.

Gamma and color correction as part of the debayer are the only reason we put up with files that need a red rocket or insane CPU power to play in realtime. Non destructive correction means changing color temperature 100 times with no image degradation.

Which has zero relations to the matter discussed here.

This does something else additionally and further tweaks the signal.

Synthesis = zero points

The color science going in to the color and gamma tranforms that are part of the SDK and the debayer are iterations that have evolved since the red one

Which has nothing to do with this.

Synthesis = zero points

Hrvoje sounds like he's describing IPP2 http://www.postmagazine.com/Press-C...introduces-image-processing-enhancements.aspx

That may be caused by your current inability to make a distinction between what you think this is and what this actually is.

It is far pereferable to use red color science in a non destructive way for this purpose.

I just posted a set of examples for Dragon showing the improvements of the signal, on top of base color science.

Good luck digging out that skin palette out of the signal.

In my opinion LUTs have a very important place as artistic or look LUTs as a easy way to access what is essentially a saved correction. There is no issue with shaping color with LUT's in a destructive way as there is no issue with shaping color with lift gamm gain. On the other hand putting a LUT out like this and claiming that it is a camera "improver" is disingenuous at best and reveals a basic misunderstanding about the process.

These improve the final image coming out of the camera by correcting the fixed and measurable micro offsets each camera and it's base mapping make.

Thank you for the valorization of my statements you have repetitive issues understanding so far.

Once the image has been debayered there is no improving, only changing and every change this "primer" makes has a consequence.

I see you have a problem with phrasing now.

Assimilation of new things is not easy and analytical part of mind tends to grab abstract-compliant straws to hold on to the safety of the comfort zone.

Just on the face of it the before and after corrections we see saturated blue skies in some images and overly magenta skin in others, there's a connection there...

Yes, the consistent connection here is your guesswork of what you "think" the image should look like, with zero reference to what you are judging, while never actually having tested the camera.

Explained another way LUTs are useful as part of a overall correcriron pushing towards a particular look. Grading on top of a corrector LUT (coming from a image that started with great color science) is a very bad idea as the LUT has limited color with it's correction and changed it in a global way non specific to the shot at hand.

I see you still have problems by translating your experience with the form to something you've never actually used yourself.

Basically you are trading red color science for Hvroje color science.

Complete nonsense.

Logic = F
Name spelling = F
Reading the thread = F
Synthesis = F

Hvroje says that the change has eliminated image color contamination making for easier correction downstream and optimised the image,

No, that is your strawman jumping out again.

Which points to the conclusion that you are obviously clueless about the distinction between light-affected color casts which camera records (which you compensate in post) and in-camera deviation of the signal (good luck compensating that in post).

Nth time:
Primers correct the fixed and measurable camera + base mapping originated signal deviations. To a possible extent.

Name spelling still presents the obstacle, I see.

To sum up: All luts are destructive.

To sum up: All image modifications tools are destructive and Aharon is stuck in a loop.

They are best used as part of a look

For Aharon.

In which case they are of course not destructive so that is cool since not related to this project.

Because of course someone designing that look had more insight about camera signal and its properties than the person actually doing detailed camera analysis.

and have little utility at the base correction stage beyond correcting for gross color casts or matching cameras or displays.

So you missed a couple of sample images then.
And basically everything I wrote.

Oh well...

AS they correct they change the image in a destructive "locked in" way and as such are not well suited to correction which is a flexible process that adresses each shot individually.

Last attempt of breaking illusions:

1) Each camera's base colour science has limitations, due to principles it relies on.

2) Each colour grading tool has limitations, due to principles it relies on.

3) These address issues those two have trouble addressing, due to principles these rely on, designed to offer the best effect possible within their limitations and limitations of the signal they shape.

4) Using a synthesis of all three and harvesting joined potential of each image transformation stage and its principles one can achieve improved imaging results and save precious time on this planet for more creative things. Tested on many cameras and used daily by many for the last year.



...



I have zero intent of convincing you or anyone to stop pushing the stone up the hill over and over again.
I can only offer what I made and can help others.

You are completely free to go grade yellow people or brown people or half yellow half brown people or magenta people and attempt to re-invent in-between skin tonalities squished in the signal by pushing and pulling in a grading suite, and many other things these achieve, and do it each and every time before starting the creative work, with zero reference and using tools which are less "destructive". I have zero issues with that. I have other priorities and routes and I'll get back to grading this feature I'm working on now, shot on Alexa, by using and testing Primers I recently made for it. By measuring its sensor and signal mapping. I got lucky again with all this "subjective" and "destructive" stuff.

All cool.
Life goes on and to each his own.

 

[Aharon Rothschild]

Hrvoje, thanks for the discussion, and I hope this back and forth clarified things for other users. Best of success with your LUT and as I don't have anything to add I won't comment further.