Brand specific colour

[NigelC]

Not sure if this is the right place to ask the question. One frequently reads about a certain brand having a particular strength in "colour science"; for example pleasing colours is often cited as reason for popularity of Canon with wedding/event photographers - likewise Hasselblad processed through Phocus. Although I have owned Canon FF in the past and have found colours to my taste that could be influenced by lens choice, i.e. Canon L lenses - a Zeiss 35/2 had warmer colours, a Zeiss 21/2.8 somewhat colder/"blue" but then these things can so easily be influences by a variety of factors such as WB settings. Slightly more recently I have found Olympus (OMD5 mk2) jpegs much nicer than those from Panasonic Lumix (GH2).

So is there anything fixed about colour output or, in RAW, can you make Panny look like Olympus with a custom profile in raw processing - is it fixed with jpegs?

[Alexey.Danilchenko]

I think if you will try to eradicate variables from all of those you will find out that there is not much difference nowadays between brands. People frequently tend to compare incomparable whilst forgetting influencing factors like lenses, filters, camera processing pipeline, flare, profiles and processing software.

[TonyW]

I agree with Alexey's comments, my experience indicating that the 'pleasing, sharper, higher resolution etc.' often cited by the brand loyal are no more than the effects of image manipulation of raw data by the manufacturer to get their corporate vision accepted. 

If your camera is supported then you may find work has already been done to mimic the look of the in camera JPEG settings for raw in your editor of choice, but you will need to work on the image data if you are trying to get a look different to the manufacturers look e.g. to make Canon images look like Nikon standards

It should be possible working with raw data to mimic the look you require and set as a preset in your editor of choice.  Working with SOOC JPEG may also be possible to tailor to your needs if the camera software allows it.

[MauriceRR]

There is no more money to developp colorscience and color vision in straight cameras market.
Today, sSmartphone cameras have much more sophisticated WB algorithm, tonemapping, hdr rendering, pleasant luts, face exposure, multiframe stacking, etc.

[BobShaw]

My experience (with Canon 5D2) was that if you bought 10 cameras of the same brand and model of consumer camera  that they would be all over the place as far as colour.
Hopefully later ones are better but I doubt it. You need to calibrate them.
That is part of the reason that the top end ones cost more.
Hasselblad colour is certainly awesome.

[degrub]

There is no more money to developp colorscience and color vision in straight cameras market.
Today, sSmartphone cameras have much more sophisticated WB algorithm, tonemapping, hdr rendering, pleasant luts, face exposure, multiframe stacking, etc.

Any basis for that opinion ?
Just curious about the science as many here work from raw and not camera jpgs which is what the statement seems to opine.

[Jack Hogan]

The science is not dedicated to a specific segment of the market: when a new algorithm is developed, if it works in one it works in the others.  If it's a good idea it is just a matter of time before it is ported to other formats.

There are currently two main drivers that I see to the next stage of picture taking called computational photography: instantaneous multiple captures and Machine Learning (or AI for short).  The first has been around forever and allows additional flexibility in assembling the pixels in the final image, yielding advantages in resolution, noise and/or DR.  The second is relatively new and automates a number of more or less difficult tasks that once had to be performed manually by skilled operators, particularly useful for unskilled point and shooters.  They both get around major weaknesses of smartphone cameras, which is why we may see some of this there first.

On the other hand any good results obtained in fast iterating products will for sure cross pollinate the serious photography market sooner rather than later, witness the slew of AI enhancements in commercial converters/plugins.

Jack

[Doug Gray]

The science is not dedicated to a specific segment of the market: when a new algorithm is developed, if it works in one it works in the others.  If it's a good idea it is just a matter of time before it is ported to other formats.

There are currently two main drivers that I see to the next stage of picture taking called computational photography: instantaneous multiple captures and Machine Learning (or AI for short).  The first has been around forever and allows additional flexibility in assembling the pixels in the final image, yielding advantages in resolution, noise and/or DR.  The second is relatively new and automates a number of more or less difficult tasks that once had to be performed manually by skilled operators, particularly useful for unskilled point and shooters.  They both get around major weaknesses of smartphone cameras, which is why we may see some of this there first.

On the other hand any good results obtained in fast iterating products will for sure cross pollinate the serious photography market sooner rather than later, witness the slew of AI enhancements in commercial converters/plugins.

Jack

Well stated.

[Jack Hogan]

Thanks Doug.

[MauriceRR]

They both get around major weaknesses of smartphone cameras, which is why we may see some of this there first.

On the other hand any good results obtained in fast iterating products will for sure cross pollinate the serious photography market sooner rather than later, witness the slew of AI enhancements in commercial converters/plugins.

Jack

Not so sure.
There is a much much bigger market in smartphone than in straight camera market.
Smartphones are limited by hardware, so there is a real need for software here, and camera in smartphone is near the only point where manufacturer can make the difference from others today.
Even if sensor are better and better in full frame camera, colors are quite the same from 20years.
I can take just à small exemple : even if Adobe made recently a subtil change in their pipeline for profiles, it's much or less the same result than in cameraraw 1...
Try a Samsung S9+ and a iPhonex, their rendering can be very different.
Of course I speak about jpg. Raw files are not rendered, they are just datas.

[Jack Hogan]

Even if sensor are better and better in full frame camera, colors are quite the same from 20years.
I can take just à small exemple : even if Adobe made recently a subtil change in their pipeline for profiles, it's much or less the same result than in cameraraw 1...
Try a Samsung S9+ and a iPhonex, their rendering can be very different.
Of course I speak about jpg. Raw files are not rendered, they are just datas.

There are hardware and then there are software 'color science' capabilities (to address the OP's question).  Differences on the hardware side are well known and limited, at least between decent phones and ILCs.  That's not surprising since sensors are made by a handful of manufacturers with a handful of sophisticated, time tested technologies.  Improvements are incremental.

On the software side it sounds like you are referring to different 'looks' in jpeg rendition engines.  Those are subjective and can vary wildly, just like the renditions of the same subject by Picasso and Vermeer would.  Changes tend to be arbitrary and driven by the fashion of the moment (you know, overly sharpened, overly saturated, HDR etc.).  The underlying processing technologies are mostly well known, with the few provisos mentioned earlier.

I am referring to something different, processing ability to render objectively better colorimetric images.  Differences between converters with standard settings here are quite limited, disruptive improvements coming along very infrequently.  When they do come along you can often find a seminal paper in a recent Conference/Symposium.  From there the new technology gets implemented in the various platforms at different speeds.

Clearly the faster iterating products have more of a chance to incorporate it sooner, as exemplified by Goggle's Pixel phones for instance.

Jack

[MauriceRR]

I am referring to something different, processing ability to render objectively better colorimetric images.  Differences between converters with standard settings here are quite limited, disruptive improvements coming along very infrequently.  When they do come along you can often find a seminal paper in a recent Conference/Symposium.  From there the new technology gets implemented in the various platforms at different speeds.

I would be glad to hear more about it.
I don't really understand what you are referring to. Color management ? Color rendering ? Color appearance models ?

[stamper]

I would be glad to hear more about it.
I don't really understand what you are referring to. Color management ? Color rendering ? Color appearance models ?

I think Google is your best friend.

[Doug Gray]

An entertaining blog that's mostly about color printing science is "John the Math Guy"

http://johnthemathguy.blogspot.com/2018/08/state-of-art-in-extended-gamut-printing_21.html

Sense of humor helps if stuff that's strictly technical puts your feet to sleep.

[Jack Hogan]

I would be glad to hear more about it.
I don't really understand what you are referring to. Color management ? Color rendering ? Color appearance models ?

Sure.  Just think about it: What disruptive changes have occurred in color science since the time that smartphones first appeared on the scene a decade ago?

The underlying theory was developed in the 19th and early part of the 20th century.  Color information captured by most of our sensors today is based on a Bayer pattern developed in the mid '70s.

There has been some experimentation but on the hardware side of 'color' over the last decades there has really been just minor tweaking of spectral sensitivity functions, mostly trading off one desirable property for another.  Is color information being captured today better than it was a decade ago?  Have a walk on the wild side with the CCD guys ;-)

On the processing side, we are still mostly using color matching functions developed 90 years ago (the 'new' and improved 2006 CIE ones are based on experimental data collected in 1956).  CAMs used today were developed over twenty years ago and are tweaked versions of earlier efforts.  Do you know what demosaicer your raw converter of choice uses?  The one in your phone?  Does it matter?  Mostly not, because wildly different state of the art algorithms have been really hard to tell apart for quite a while.

To find substantial differences we need to close the objective color science chapter and open up the perceptual image processing one (skipping the artistically pleasing but unrelated to science one).  Most of these are old(er) techniques that are finally coming of age because of processing power that was not available 20 years ago.  For instance floating point vs integer math, more advanced profiles, wavelets, etc.   More powerful computing power also opens up the way to computational imaging, as discussed earlier (in-camera or in-computer).

So the reason that by-and-large LR Process Version 2003, 2010, 2012 and Current all look so similar is that the underlying color science has pretty well remained the same.  And the reasons why many phone images look the way they do instead is because manufacturers (not all) are trying to appeal to the Instagram crowd, whose tastes change faster than you can say cool-aid.

Jack

[DP]

I can take just à small exemple : even if Adobe made recently a subtil change in their pipeline for profiles, it's much or less the same result than in cameraraw 1...

nope... you will really notice changes between simple matrix profiles vs what we have now in many color tones... do you want a simple test ? Adobe still hardcodes simple matrix profiles (cameraraw 1) in the code itself... remove all external dcp profiles (Adobe Standard, etc) for a given camera model, do conversion... return, do conversion... check orange/red tones for example...

[MauriceRR]

Ok right, they used now kind of tweaks for color memories. But what about color models for hdr vs ldr ? What about colors and surounding factors ? What about tonemapping ?
Color for photography is still at the stone age.
Even the video side of cameras is a lot more developped for colors, tone compression, etc. If you look a software like 3dlutcreator, which is well suited for photography too but comes from the vidéo world, this soft is more attractive in terms of color parameters than Adobe cameraraw and photoshop, or more interesting than standards jpg ooc from camera manufacturers. Lot of things in terms of colors and tone compression, especially in Adobe, is kind of à Black box for photographers, and it's a shame. It's very difficult to do retro engineering with a tool from Adobe cameraraw process v4 and dcp profiles.

[ErikKaffehr]

Hi,

I would assume that there are two major differences:

• White balance
• Camera profiles

Getting a correct white balance is instrumental. So, if you have a camera that has a good auto white balance, that may help a lot.

The color coming out of a camera is essentially decided by the camera profile.

The attached image was shot with three very different cameras I have:

• Phase One P45+ (Kodak CCD back, released in 2007)
• Sony A7rII (Sony CMOS, released in 2015)
• Sony A900 (Sony CMOS, released in 2008)

All three were processed with Lightroom, using a DCP profile for each camera that was based on the color checker in the image. Such a profile was generated for each camera, using LumaRiver Profile Designer.

I don't know if you can see a significant difference.

LumaRiver Profile designer is math-based. It finds an compromise matrix based on the ColorChecker and combines it with a 'look'.

Color profiles used in Capture One and Hasselblad's focus are probably different, it is said that Capture One's conversion is not matrix based. But, both are probably hand made. It is said that Hasselblad's 'Phocus' produces very similar colors, although the cameras can have different sensors.

Would you be really interested in those images, a layered TIF is here: http://echophoto.dnsalias.net/ekr/Articles/Temp/Tricolore/Tricolore2.tif

That layered image also has a layer with colors sampled from the fruits, measured with a spectrometer.

Best regards
Erik

Not sure if this is the right place to ask the question. One frequently reads about a certain brand having a particular strength in "colour science"; for example pleasing colours is often cited as reason for popularity of Canon with wedding/event photographers - likewise Hasselblad processed through Phocus. Although I have owned Canon FF in the past and have found colours to my taste that could be influenced by lens choice, i.e. Canon L lenses - a Zeiss 35/2 had warmer colours, a Zeiss 21/2.8 somewhat colder/"blue" but then these things can so easily be influences by a variety of factors such as WB settings. Slightly more recently I have found Olympus (OMD5 mk2) jpegs much nicer than those from Panasonic Lumix (GH2).

So is there anything fixed about colour output or, in RAW, can you make Panny look like Olympus with a custom profile in raw processing - is it fixed with jpegs?

[Jack Hogan]

Ok right, they used now kind of tweaks for color memories. But what about color models for hdr vs ldr ? What about colors and surounding factors ? What about tonemapping ?

Yes, disruptive improvements in those are the kind that you read about in the proceedings of an obscure symposium and a few months/years later trickle into the digital imaging processing chain - any camera processing chain, independent of format.

Color for photography is still at the stone age.

When you say that, it appears that the meaning you are ascribing to terms like 'color' and 'color science' is different than what I am using.  The difference is that 'color science' attempts to produce quantitative and objectively 'accurate' colors, taking perceptual effects into consideration ...

Even the video side of cameras is a lot more developped for colors, tone compression, etc. If you look a software like 3dlutcreator, which is well suited for photography too but comes from the vidéo world, this soft is more attractive in terms of color parameters than Adobe cameraraw and photoshop, or more interesting than standards jpg ooc from camera manufacturers. Lot of things in terms of colors and tone compression, especially in Adobe, is kind of à Black box for photographers, and it's a shame. It's very difficult to do retro engineering with a tool from Adobe cameraraw process v4 and dcp profiles.

... while it seems that you are referring to producing qualitatively interesting, subjectively 'pleasing' tones.  Nothing wrong with that but the deltaE00 coming out of those two definitions is huge.

Jack