I would like to know wich camera system offer that nice color

[ErikKaffehr]

Hi Edmund,

As far as I recall, you have pointed me to "Dietmar's" device. Regarding the Passport I feel it is quite useful.

Best regards
Erik

Erik,

 If it's relevant, I'm aware of Dietmar's device, and obviously I'm aware of the Passport as I was involved in its design
 It's late and I'm not thinking straight, but I believe that once you have the sensor's spectral sensitivity curve (eg. Dietmar's gadget), you still need to measure the illuminant in situ with a spectro. Even if you scan your scene with a teleradiometer or an XX channel camera,  to turn the result into color you still need information about the illuminant

Edmund

[BernardLanguillier]

Another scenario is in skin colors. I guess you are familiar with the way c41 films render skins (portra 160 for example). I'm talking about this slightly over exposed skin look. Well, try to do the same on a d800, there will be no detail, the skin seems like one big blub that only get's sort of details when it goes into the left side of the histogram. I'm not talking about recovering highlights (which is also pretty terrible, but that's where film wins over both digital backs and cmos), but about this strange behaviour where the subtilities of lighter skin can't be represented by the D800, but both C41 film and some samples I've seen of digital backs have no problem to represent it.

Is this any close to the kind of look you are talking about?



Cheers,
Bernard

[jerome_m]

You should start a company that provides that service for any Digital camera owner. One sends in the camera, you do the custom calibration, solve the linearisation issues and write a few files and voilá, great MFDB like color in any DSLR (or mirrorless) camera. Im sure you can make good money.

For this to be mathematically possible, cameras would need to fulfil a set of conditions called the Luther-Ives conditions. As a general rule, they don't.

[ErikKaffehr]

Hi,

Well, the Luther-Ives conditions are impossible to fulfil, that would require negative sensibility.

The articles below gives some good insight:

http://dougkerr.net/Pumpkin/articles/Metameric_Error.pdf
http://dougkerr.net/Pumpkin/articles/Sensor_Colorimetry.pdf


Best regards
Erik

For this to be mathematically possible, cameras would need to fulfil a set of conditions called the Luther-Ives conditions. As a general rule, they don't.

[ErikKaffehr]

Hi,

What surprises me a bit that a great majority of competition winners are shot with inferior cameras like Canon and Nikon used by superior photographers.

Here is a good example, enjoy: http://yanzhangphotography.com

On the other hand, owning an MFDB doesn't make you a mediocre photographer either! Some examples:

• Ed Hurst's images impress me and he is using MFD creatively: http://www.spiffingpics.com
• Wayne Fox is a frequent poster on these threads, always offering good advice, you find him here: http://www.landscapesbywaynefox.com

This is a thread with a lot of good and a lot of not so good MF images:
http://www.getdpi.com/forum/medium-format-systems-digital-backs/4730-fun-mf-images.html

Best regards
Erik

Is any close to the kind of look you are talking about?



Cheers,
Bernard

[Bart_van_der_Wolf]

When you raise or lower the shadows it seems to only affect the shadows, if you do the same on the D800 you will get halos around every detail, and the color quality goes down the drain dramaticly. There seems to be some sort of inertia spanning over larger areas (dozens of pixels) which gives this sort of tonemapping look. I'm sorry I don't have any technical term for this, but I think you all know what I'm talking about.

Hi,

What you are describing is the effect of poor Raw conversion/tonemapping. It has nothing to do with the camera itself.

Trying to fit the full dynamic range of a D800 in the limited output DR leads to a dull image, therefore the conversion software has to apply tonecurve adjustments which, when not done well, lead to the type of effects you describe. I think it is more prominent in LR/ACR types of Raw conversion where a lot goes on under the hood, especially in highlight compression.

Another scenario is in skin colors.
[...]
I'm not talking about recovering highlights (which is also pretty terrible, but that's where film wins over both digital backs and cmos), but about this strange behaviour where the subtilities of lighter skin can't be represented by the D800, but both C41 film and some samples I've seen of digital backs have no problem to represent it.

Again, postprocessing. Try using a more linear tonemapping curve (which would require at least a lot of Highlight slider reduction in Adobe's Process Version 2012 conversions, which doesn't recover but compress highlights), followed by tools like Topaz Clarity (for local halo free contrast adjustments at several levels of original captured contrast) and Detail (for skin/hair/clothing detail). This type of behavior is much more benign in a converter like Capture One, which also has very useful skin tone correction capabilities, allows to create one's own tweaked version of Profiles, and offers a better starting point for further postprocessing ...

Cheers,
Bart

[Bart_van_der_Wolf]

What surprises me a bit that a great majority of competition winners are shot with inferior cameras like Canon and Nikon used by superior photographers.

Hi Erik,

What do you think the criteria of the judges are, who is the organizer, ...
That's one of the reasons why over the top HDR tonemapping is popular.

And, given two identical situation shots (same file), one properly processed and one not, which do you think would win ...
And, define properly processed if we go beyond the mere butchering with halo ridden tonemapping ...

Cheers,
Bart

[Chris Livsey]

What surprises me a bit that a great majority of competition winners are shot with inferior cameras like Canon and Nikon used by superior photographers.

Here is a good example, enjoy: http://yanzhangphotography.com

Best regards
Erik

Are you sure that's not cgi?

[ErikKaffehr]

Hi,

I just went to Photo.net and picked one of the highly ranked photographers, he is using Nikon.

On the other hand I may agree that we have a trend to over processing images, but it sometimes take some processing to match the perception of a scene.

In B&W times we used orange/red filters, in the slide era it was Velvia and Polarizers. Another point is that the cameras have a very wide dynamic range, with essentially no tonal compression (like toe and shoulder in the film times), so we need to supply that tonal compression to display on devices with limited luminance range.

These image of mine is probably a bit overprocessed, but I like it anyway:


Best regards
Erik

Are you sure that's not cgi?

[jerome_m]

Well, the Luther-Ives conditions are impossible to fulfil, that would require negative sensibility.

Of course not, since human eyes fulfil them, by definition.

[ErikKaffehr]

Now, I don't think so, it is the human eye in combination with signal processing in the brain.

Best regards
Erik

Of course not, since human eyes fulfil them, by definition.

[Bart_van_der_Wolf]

On the other hand I may agree that we have a trend to over processing images, but it sometimes take some processing to match the perception of a scene.

Hi Erik,

Post-processing is inevitable, because it is impossible to render the captured dynamic range (11-14 stops of engineering DR with an overlay of veiling glare, noise reduction can tame lots of the shadow noise) on a 7-stop output medium without it looking extremely dull and flat.

Tonemapping is unavoidable, and also offers a great opportunity to shape the tonecurve to match out creative vision as good as possible. It requires a lot of experience and hard work, or great tools (like Topaz Clarity) to assist us. It's still the photographer who makes the difference, also in postprocessing.

In B&W times we used orange/red filters, in the slide era it was Velvia and Polarizers. Another point is that the cameras have a very wide dynamic range, with essentially no tonal compression (like toe and shoulder in the film times), so we need to supply that tonal compression to display on devices with limited luminance range.

Exactly. We did part of the tonemapping in camera, at capture time, because that was the relatively easy way. Those with darkroom capabilities could enhance that by tuned film processing, and tuned printing techniques (dodging/burning/local developer temperature/filters). Nowadays we have much better tools at our disposal, and in color.

Cheers,
Bart

[ErikKaffehr]

Hi Jerome,

Not really, the response function of human vision is a combination stimuli and processing in the brain.

Best regards
Erik

Of course not, since human eyes fulfil them, by definition.

[Chris Livsey]

Hi,

I just went to Photo.net and picked one of the highly ranked photographers, he is using Nikon.

On the other hand I may agree that we have a trend to over processing images, but it sometimes take some processing to match the perception of a scene.


These image of mine is probably a bit overprocessed, but I like it anyway:


Best regards
Erik

Erik, your shot looks like a photograph, that is a compliment BTW, some work looks like cgi, that doesn't mean it is without merit and may indeed win a competition (depending on the criteria and the judges, whatever they are worth).

[Jim Kasson]

Hi Jerome,

Not really, the response function of human vision is a combination stimuli and processing in the brain.

Erik, I side with Jerome on this one. As an example of a set of everywhere-positive response curves that is a 3x3 matrix multiplication away from CIE XYZ, I cite the CIE 1931 Standard Observer.  

http://en.wikipedia.org/wiki/CIE_1931_color_space#CIE_standard_observer

In fact, though I can't prove it (darn, I wish I'd taken linear algebra -- Math 124 -- at Stanford), I strongly suspect that, just as there are an infinite number of linear transforms of XYZ, there are an infinite number of linear transforms of XYZ that are everywhere positive.

Jim

[synn]

This straw man argument about contest winners that came out of nowhere is hilarious. What is the install base of canikon users vs MFDB? Of course you will see a disproportionately high number of canikon shooters participating in contests and subsequently winning them. Does that make them "better"?

Beyoncé has more Grammys than yo-yo ma. Does that make her music better?

Look at the proportion of canikon vs MFDB in color critical applications like repro. What do you see there?

To some, everything is a science. Even art. Fair enough. Then please, someone create a color profile for say, a D800 that under any shooting circumstance, deliver an image that looks exactly like it came from a phaleafblad back. I am willing to contribute as many RAW files as needed for this. Even if you get every test chart to look exactly the same, I can guarantee that the real world images would not.

Go ahead, surprise me.

[ErikKaffehr]

Chris,

I can freely admit my example was not a good one. My take on the issue is that I would guess something like 60-85% percent of the leading photographers use Nikon or Canon.

If there is some collection of the worlds best images, probably something like 70% of them would be shot with a Canon and a lot of them would be with shot with the 16-35/2.8L.

I don't keep record of competitions. But lets look:

- Josh Holko, a frequent poster and also a frequent winner of competitions uses Canon: http://blog.jholko.com/equipment-2/
- Jack Dykinga uses Nikon and 4x5" film
- Frans Lantink uses Nikon: http://www.lanting.com/camera-gear.php
- Hans Strand uses Hasselblad and Nikon: http://www.hansstrand.com/Hans_Strand/Equipment.html
- Andy Biggs, also well known on these forums uses Phase One, nowdays: http://andybiggs.squarespace.com/the-global-photographer/category/photo-gear-and-reviews
- Mattias Klum is using Nikon
- Joe Cornish is using Panasonic G3, Nikon D700/D800 but his main camera is Phase One IQ 180.

So, quite clearly some of the greatest photographers are happy with Canon and Nikon, but folks on these forums often say those cameras have crappy rendition.

Personally, I use a not so old P45+ on an old Hasselblad and 2-3 Sony DSLRs, and I am happy with both, but I am no great photographer :-)

Best regards
Erik

Erik, your shot looks like a photograph, that is a compliment BTW, some work looks like cgi, that doesn't mean it is without merit and may indeed win a competition (depending on the criteria and the judges, whatever they are worth).

[ErikKaffehr]

Hi Jim,

I was mostly thinking about this figure , but also some articles mentioning that negative sensivity is required to reproduce human vision.

Best regards
Erik

Erik, I side with Jerome on this one. As an example of a set of everywhere-positive response curves that is a 3x3 matrix multiplication away from CIE XYZ, I cite the CIE 1931 Standard Observer.  

http://en.wikipedia.org/wiki/CIE_1931_color_space#CIE_standard_observer

In fact, though I can't prove it (darn, I wish I'd taken linear algebra -- Math 124 -- at Stanford), I strongly suspect that, just as there are an infinite number of linear transforms of XYZ, there are an infinite number of linear transforms of XYZ that are everywhere positive.

Jim

[Jim Kasson]

I was mostly thinking about this figure , but also some articles mentioning that negative sensivity is required to reproduce human vision.

Erik, those aren't spectral responses per se. Those are the CIE 1931 RGB Color matching functions. The color matching functions are the amounts of primaries needed to match the monochromatic test primary at the wavelength shown on the horizontal scale in the classic color matching experiment.

Consider the normalized cone responses as another example:



Does that make sense?

Jim

[ErikKaffehr]

Hi Jim,

I was thinking about this passage from Doug Kerr's article:

A Luther-Ives sRGB sensor
Imagine that we had a sensor that:
• Met the Luther-Ives conditions, so that any two instances of light
having the same color, even with different spectrums, would be
reported with the same output values.
• Had outputs would be the (linear) coordinates of the sRGB color
space.
To meet the Luther-Ives conditions, the response curves of all three
channels of this sensor would have to be (different) linear
combinations of the eye cone response curves, l, m, and s. We can in
fact determine the “coefficients” of that combination (the constants
by which one, two, of three of the cone response curves were to be
multiplied before they were added to give each of our response
curves) with a straightforward mathematical process.
So, why don’t we just build such a sensor? It would be so handy.
There would be no need to transform its outputs into linear sRGB
coordinates—they already are.
Aye, but here’s the rub. Such a sensor’s response curves would have
to have negative values in parts of the wavelength range, which is of
course impossible to implement in a real sensor (it is not even
physically meaningful). Figure 4 shows these response curves (again,
labeled with the names of the primaries, not the names of the curves).
Figure 4. sRGB response curves
So we must consider another sensor design for our real camera.

Best regards
Erik

Erik, I side with Jerome on this one. As an example of a set of everywhere-positive response curves that is a 3x3 matrix multiplication away from CIE XYZ, I cite the CIE 1931 Standard Observer.  

http://en.wikipedia.org/wiki/CIE_1931_color_space#CIE_standard_observer

In fact, though I can't prove it (darn, I wish I'd taken linear algebra -- Math 124 -- at Stanford), I strongly suspect that, just as there are an infinite number of linear transforms of XYZ, there are an infinite number of linear transforms of XYZ that are everywhere positive.

Jim