Improving the Camera: The Image

[Wayne Morellini]

The age of the viewer can also affect its tinge, as happened with I forget the impressionists name, who repainted a scene decades latter to find the colors changed.

 

[Hrvoje Simic]

Yet I have often seen the color of the sky, having also regularly looked at the sky to forecast weather

Wayne...there is no single color of the sky.

Sky color varies depending on:

- time of day
- angle of Sun rays and resulting atmosphere light polarization,
- longitude and lattitude
- atmosphere conditions (pollution, weather)
etc...


I cannot convince you what was the colour of the sky in those shots nor I can target what you expect to see.
I can however reduce the signal offsets so the camera gives more referent imagery.

But one can reproduce the exact spot checked color and most of the time be wrong, as a scene depends in the dynamic interplay with the accommodation of human visual system and light sources and temperature.

Color of the scene depends on light.

The interpretation of the scene depends on human visual system, reference and conditioning/bias.

Designing a good camera accommodates to making the scene more neutral to the color seen, rather than the color measured.

"Color seen" is a moving target.

Color response can be measured for reference and when that is optimally applied taking many things into account it leads to "color seen". Minus the occasional particular subjective bias.

If you shift everything to measured generically, it will look unnatural, instead of how the scene looks as the starting point.

Generic image transformations are done that way, targeting primaries, gamma and gamut and else as it falls.

This is not achieved generically.
Exactly the opposite. Taking into account sensor response and signal particulars.

 

[Wayne Morellini]

Wayne...there is no single color of the sky.

Sky color varies depending on:

- time of day
- angle of Sun rays and resulting atmosphere light polarization,
- longitude and lattitude
- atmosphere conditions (pollution, weather)
etc...


I cannot convince you what was the colour of the sky in those shots nor I can target what you expect to see.
I can however reduce the signal offsets so the camera gives more referent imagery.



Color of the scene depends on light.

The interpretation of the scene depends on human visual system, reference and conditioning/bias.



"Color seen" is a moving target.

Color response can be measured for reference and when that is optimally applied taking many things into account it leads to "color seen". Minus the occasional particular subjective bias.



Generic image transformations are done that way, targeting primaries, gamma and gamut and else as it falls.

This is not achieved generically.
Exactly the opposite. Taking into account sensor response and signal particulars.

Hrvoje. When somebody says they are a regular sky watcher, don't you think it implies they are familiar with all the colors of the sky under a number of conditions.

However, that is my point, certain cameras try to normalise the image from the digital domain to how the normal eye perceives it. While you can offset deliberate color imbalance, you can eventually land up fighting the free normalisation as well.

Yes, that is what I am saying it depends on "human visual system, reference and conditioning/bias".

Color seen is a moving target, but in the elongated moment it is often a roughly commonly perceived elongated moment. In such, as it often does not depend on momentary differences, and sometimes shifted difference in visual response between two nearby observers under the same conditioning, as much as presenting a close reality of what a single observer on average may see to the public.

So, again, the scientific color measured, and the camera systems conditioning of the image closer to how a person would see those colors and contrast range in that instance, minus, if it exists on what ever camera, any possible camera image distortion to get somebody to buy a more expensive camera.

"Generically" as in taking into account measured colors, rather than how the visual system conditioning would perceives them in the elongated moment which cheaper cameras bake into the image, which as baked, leads to loss as people fight it, not being RAW to start with.

 

[Mike Nagel]

Hrvoje,

I need u to be perfectly honest.

Are u in any way associated with this piece ?!

The Windmill Massacre (2016)

 

[Hrvoje Simic]

Helium Beta nearing completion.



Images untouched, only applied two different routes from Log3G10/RWG to 709.

- First one Primers designed for Helium response,
- Second one IPP2 low cont 2 LUT.

Same ISO & WB.





Erich Roland, DC Camera


Primer

Primer

Primer

Primer

Primer

Primer

Primer

Jason Han


Primer

Primer

Primer

Beta testing > exert at omeneo.

 

[Joel Arvidsson]

Im ready when you ready! Looking forward for the release version!

 

[Wayne Morellini]

Untouched, Helium ungraded. I thought those pictures looked good.

 

[Craig Parkes]

And those shots where such an approach may or may not fall down. The blues in the sky that have been selectively boosted by the primer have been boosted quite significantly to what I would expect.

The plus side is if I wanted to knock them back and wanted to also keep a particularly vibrant blue look on something in the foreground of the shot that is that it's less noticeable and less work to dropping a mask on the sky normally to knock things back then it is to figure out if you want to boost the blue in the whole scene or just in a colour range (and by the nature of LUTs vs Keys the LUT is obviously targetting things more specifically)

In this way the name Primer hold's true as in being a good firsy step, but may be slightly inappropriate analogy given a primer in the painting medium is something that you would do and add to (in totality or selectively) to get to the desired final result. In an electronic sense a primer may be somethinb that may get you pretty close but then you want to correct away from some parts.

Which I think does speak a little to one issue of using a LUT to try and give them image a desired look early on, the workflows available don't allow you to selectively turn down a lut in parts (without compositing the LUTted imagery and the unlutted imagery in one shot) so the level of control you have over the imagery is decreased.

In saying that, the subjective choice in the design of the LUT could end up catering to imagery in a way where the particular elements that makeup parts of a frame that are hard to selectively boost (unpredictable moving objects like human faces) are inherent closer to where you want them. Whereas the things that look unusual/not normal are the things which tend to be easier/more predictable to correct.

 

[Wayne Morellini]

+1

Yes, the skies, and the golden hair. We should never assume the Bayer is delivering what is authentic, one they don't, and two it depends on sensor and the color mask, it may not even be close. So your neutral starting point is somewhere else, and a primer needs to be ideally generated pet scene by cards (all other the place, but at least on a electronic shutter board for reference. The software can take what it already knows and the scene reference to produce pretty neutral). Just cone up with a much better way to do it actively all the time in camera. Rather interesting solution. So, a genetic lut and a light box with color and latitude to set primer up once every year or so (as sensor/lens ages), and some camera information on shutter gain, aperture lens, to determine brightness of each point in the scene, and maybe a slapper card, and the primer can auto nestralise to a higher accuracy, as an intermediate solution Hrvoje?

 

[Hrvoje Simic]

Something for the first day of Spring.

Primers EQX for Helium are ready.

During the work on Helium I learned some things so implemented them in Dragon Primers as well.
Primers EQX for Dragon are updated. Links are in the mail.

Happy Equinox 2017...



http://omeneo.com/primers

 

[Wayne Morellini]

Any samples?

Well, I was traveling, opened my eyes one morning, and the sky looked oddly like your skies for a few seconds, but latter I discovered if I just slid open the fly screen, they looked normal again.

 

[Mike Nagel]

Any samples?

Well, I was traveling, opened my eyes one morning, and the sky looked oddly like your skies for a few seconds, but latter I discovered if I just slid open the fly screen, they looked normal again.

;-))))))))))

 

[Joel Arvidsson]

Sweet this I have been waiting for!

Something for the first day of Spring.

Primers EQX for Helium are ready.

During the work on Helium I learned some things so implemented them in Dragon Primers as well.
Primers EQX for Dragon are updated. Links are in the mail.

Happy Equinox 2017...



http://omeneo.com/primers

 

[Hrvoje Simic]

The blues in the sky that have been selectively boosted by the primer have been boosted quite significantly to what I would expect.

These are not designed to meet subjective expectations but allow more referent starting point. Subjective expectations are a moving target and best met in grading suite.

These deal with fixed targets and what can be done before the grading process.

Both routes are "boosted" from the base colour science which is flat and undersaturated to allow specific mapping and consequently shaping of data and that is normal and necessary route with log and wide gamut mappings to get to a correct image. What differs is how is what done and what is the result.

That result predetermines what you can do in post due to signal distribution you get to work with.
You cannot re-invent squished tonalities, correct very particular deviations or bring back slammed saturation in color grading process without penalties, due to inter-dependent properties of image modifying tools and color space they operate in.

These tools are not designed to be a substitute for color grading work but as an extension to allow achieving improved results when used in conjunction.

In this way the name Primer hold's true as in being a good firsy step, but may be slightly inappropriate analogy given a primer in the painting medium is something that you would do and add to (in totality or selectively) to get to the desired final result. In an electronic sense a primer may be somethinb that may get you pretty close but then you want to correct away from some parts.

Surface typically has to be treated before applying the primer.

Before applying Primers EQX, designed for raw and log based 10bit + imagery you want to prep the surface by choosing base gamma, gamut, ISO, temperature and tint in raw and (if needed ) signal distribution in RGB log. There are four main DR mappings with differing contrast, to make the process more practical.

After you apply the Primer on that basis you get a more accurate image with improved separation and full DR usable, with no data clipped or clamped, and you can do overall toning, light compensation and desired luminance distribution in log before/ inside the Primer and particular toning, final saturation boost and final brightness in video after the Primer, with a harmonized signal.

Pre-Primer part (raw + log) defines what goes into the Primer for shaping, Primer optimizes the signal in multiple aspects, and post-Primer part shapes the final outcome. Three integrated stages of post image manipulation which allow achieving more than one or two.

Also, if you use these in a monitoring path and judge exposure based on visual feedback you are less likely to underexpose and more likely to notice and avoid unwanted color casts.

In saying that, the subjective choice in the design of the LUT could end up catering to imagery in a way where the particular elements that makeup parts of a frame that are hard to selectively boost (unpredictable moving objects like human faces) are inherent closer to where you want them.

That is not what these do.

These are reference-based signal shaping tools embodied inside a LUT based conversion to viewing/delivery standard.

If this was subjective design to give "something" it would be done in one afternoon and inadequate for general application.
This takes months, custom in-depth camera profiling and massive back and forth between measurable technical accuracy and practical real-world application.

 

[Hrvoje Simic]

"My ignorance is vast though constantly decreasing."


Primers EQX for Dragon and Helium have been updated.
Links are in mails...


Have a great Sunday...

 

[Joel Arvidsson]

Hrvoje, i was going to download the updated helium primers just now but the download link had expired!
//Joel

 

[Hrvoje Simic]

No problem, check your mail.

 

[Hrvoje Simic]

Testing Primers for Ursa Minor today on a shoot with Minor Pro.
Whoa...

 

[Christoph Tilley]

Hey guys,
I did a quick review of the primers and other LUTs when grading IPP2 in Resolve 12.5:
http://www.reduser.net/forum/showth...orkflow-in-Resolve-12-5&p=1724396#post1724396

Would love to hear your feedback, and thanks for the awesome LUT, Hrvoje!
Best,
Chris

 

[Hrvoje Simic]

Further improved. Both Dragon and Helium.
Links are in mailboxes....



To stress again:

a) This all relies on Red base colour science and optimizes the result through the final stage before the color grading process

b) This is NOT creatively graded for a subjective "look" but designed based on in depth camera analysis for signal offsets, and signal is reshaped
within an image transformation step to standard gamma and color space, and its purpose is to give a more referent starting point.

c) This works across many scenes and lighting scenarios because it is NOT "scene referred" but "sensor referred" type of camera-specific image transformation and optimization.

d) Full DR, whole captured gamut compressed inside a target space.
No clipping, clamping or posterization.

e) It has multiple usage, from live monitoring and hitting the right exposure with a healthy signal and full visible DR, over practical dailies tool to a camera specific signal shaping tool for post.





Dragon

(Jacobo, Luigi)






02 NTR Primer

IPP2LowcontR2 LUT

1:1 Crops (best viewed on a large screen)




02 NTR Primer

IPP2LowcontR2 LUT

.......................................



02 NTR Primer

IPP2LowcontR2 LUT

.......................................




02 NTR Primer

IPP2LowcontR2 LUT