For Iliah Borg - D100 Characteristic Curves Test

Started Apr 18, 2004 | Discussions
fotogenetic Senior Member • Posts: 2,958
For Iliah Borg - D100 Characteristic Curves Test

Iliah,

I performed a controlled test for all the following tones ( http://fotogenetic.dearingfilm.com/figures.html ):

Normal
Less
More
White Wedding v3.3
Provia v3.2
JT all-in-one
PlanetNeil 159.

The results were interesting. They handle tone very similarly with the ones I created holding slightly more highlight details, but not by much.

I also created a density plot for each of the curves, however, I cannot directly compare these to the densities in film because I do not know how to convert from EV to Log Exposure. How do you do this? The Excel spreadsheet I used is downloadable in the link.

OP fotogenetic Senior Member • Posts: 2,958
Working link w/ graphs [imgs]
FredrikF Regular Member • Posts: 263
Re: Very small comment

Thanks for doing all this work with the curves. I've been using one of yours for a while now, and I'm very satisfied.

About the curves: It's easier to read the graphs if the same tone curve has the same colour/symbol in all figures. (The physicist in me speaking)

-- hide signature --

FredrikF

adanac Regular Member • Posts: 403
Re: Working link w/ graphs [imgs]

Question:

What is the actual procedure in making the graph?

How many images did you take for each of the plots?

Just curious!

Arjh

fotogenetic wrote:

http://fotogenetic.dearingfilm.com/figures.html

OP fotogenetic Senior Member • Posts: 2,958
Re: Working link w/ graphs [imgs]

In Ansel Adam's book "The Negative", he suggests taking a series of photographs like these before using any film to fully understand how it handles tonality. All you have to do is take a photograph of a gray card with your camera on a tripod using each of the tone settings you wish to test. I hooked up my D100 to my computer via the USB cable and took 17 photos. I spot metered on the gray card and went from -4.5 EV all the way up to 3.5 EV (Exposure Compensation).

adanac wrote:
Question:

What is the actual procedure in making the graph?

How many images did you take for each of the plots?

Just curious!

Arjh

fotogenetic wrote:

http://fotogenetic.dearingfilm.com/figures.html

Iliah Borg Forum Pro • Posts: 29,605
Re: For Iliah Borg - D100 Characteristic Curves Test

Great. Really wonderful.

I tried to overlay a plot of reproduction (target) values of RGB as a function of D for gammas 1.8 and 2.2:

  • for eV from -5 to +3.5

  • D (density) = = -LOG(exp.calibration)-eV/LOG(10;2) (exp.calibration here can be 0.12 and 0.18, for example)

  • RGB = =255/10^(D/gamma)

Not too funny... Even with 12%.

to convert from EV to Log Exposure.

It is log of relative exposure, 0.5eV = 0.16 log H
--
no text

Iliah Borg Forum Pro • Posts: 29,605
Re: Working link w/ graphs [imgs]

"Y" axis label -- D=-LOG((RGB/255)^gamma)
and maybe "brightness" is a bit ambigous here
--
no text

OP fotogenetic Senior Member • Posts: 2,958
Re: For Iliah Borg - D100 Characteristic Curves Test

Would you be able to compare the tonality of the D100 with that of Velvia or Provia using the Excel spreadsheet I created?

Iliah Borg wrote:
Great. Really wonderful.

I tried to overlay a plot of reproduction (target) values of RGB as
a function of D for gammas 1.8 and 2.2:

  • for eV from -5 to +3.5

  • D (density) = = -LOG(exp.calibration)-eV/LOG(10;2)

(exp.calibration here can be 0.12 and 0.18, for example)

  • RGB = =255/10^(D/gamma)

Not too funny... Even with 12%.

to convert from EV to Log Exposure.

It is log of relative exposure, 0.5eV = 0.16 log H
--
no text

Iliah Borg Forum Pro • Posts: 29,605
D100 Characteristic Curves Test

Banal findings:

From the curves you created it is IMHO very definite that Nikon goes with nearly 12%, and all the custom curves are applying TRC to bring 12% to about 20-22%. Shifting custom curves to the right 0.5 eV brings them to pretty good match to the "normal" curve, exept for the shoulder, which results in preserving highlights due to extra room. It is also pretty obvious that trying to keep lower noise levels, "normal" curve renders shadows darker then they should be. And thus the smooth curve renders higher gamma in midtones and highlights, compressing them. Shoulder of custom curves come very close to Gretag-Macbeth recommended rendition of highlights, having RGB=249 up to D=.23, which regular gamma 2.2 curve would render to approx. 200 This demands for a sharper contrast in midtones, rendering them more punchy. From the point of view of printing and tone separation "normal" curve is not very good. For photoprinting custom curves seem to provide more pleasant images. For offset printing... all the tone curves need extensive post-processing.

I hope I understood your question below.

Would you be able to compare the tonality of the D100 with that of
Velvia or Provia using the Excel spreadsheet I created?

-- hide signature --

no text

tony field Forum Pro • Posts: 13,802
H&D curve of film vs digital

fotogenetic wrote:

Would you be able to compare the tonality of the D100 with that of
Velvia or Provia using the Excel spreadsheet I created?

Your curves match what I did a year or so ago on the D100 and negative film material. The only problem with relation to film is to match the high density end of the curves with what you have in digital. However, they can be directly related with a simple linear equation on the straight line portion of the H&D curve. My basic proceedure was to map Zone 5 on film to the out-of-camera light meter reading of a grey card.

I first used a flat bed scanner to measure the density of the films. This proved to be less than excellent.

Subsequently, I went to the local university graphic art department that has a transmission densitomiter and measured the density of the film zone exposures (also try the Fine Arts department if they have a photography course - or maybe a very good pro printing shop). I then scaled the image density numbers so zone 5 was mapped to the same density as a digital camera's light meter reading (not quite prefect since meters are set for 12-15% reflectivity and not 18% that Zone 5 should be). Zone zero was pegged at the same digital density as digital noise (almost RGB 0,0,0 but I arbitrarily tried 0,0,0 and 1,1,1 as well).

After plotting (on paper) the curves, the net effect was that negative film had about 4 more stops of highlight information than the d100. The film curve was the classic H&D S-curve. The digital curve was essentiall a straight line with a slight compression in the highlights. (Humm, guess the electron wells can froth and foam

From this test, I concluded that the primary reason for the "beauty" of film images is that the contrast compression in the highlights and shadows allow a much finer tonal gradation than digital. Digital does not have the fine modulation in the shadows or highlights. The worst case is to photograph the sun in the picture and see the concentric rings of reddish tonality on digital that is represented as smooth tonality on film.

Like you, my custom curves have slight compression in the highlights to mimic film's preservation of highlight. I had not considered adjusting the curves midpoint as you did.

I basically concluded that I could directly relate the densitometry readings with digtal by using a simple factor - but I don't have those notes since I moved.

To define the exact shape of the digital vs film curve at hight and low intensities, I photographed a white 16x20 card that was lit with 45 degrees (relative to the card plane) incident light and the light was about 2 inches away from the leading edge of the card. This gave a continuous gray scale because of inverse square law. I chose zone 2 and zone 11 exposures using my Minolta one degree spot meter for the leading edge (bright side) of the images and and attempted to relate the digital and film captures. This did little more than confirm that digital was essential a "linear" log density function and that film had the classic S-curve as expected. This "proved" that digital cannot hold shadow or highlight texture. I vaguely recollect that I should have chosen zone 3 for the dark end test point to ensure that I would be on the linear portion of the film H&D curve.

Any digital experimentation with a custom full S-curve (both toe and heel comrpessions) proved useless - probably due to the lack of dynamic range of the digital camera. The only meaninful gain was in highlight compression.

tony

tony field Forum Pro • Posts: 13,802
forgot to mention...

fotogenetic wrote:

Would you be able to compare the tonality of the D100 with that of
Velvia or Provia using the Excel spreadsheet I created?

If you run your test curves with the three contrast settings of the D100, you might find that this effectively results in Ansel Adam's normal minus, normal, and normal plus developement. This is great if you actually use a spot meter (in camera or external).

I don't know if this attempt to quantify the behaviour of the camer system "really helps photography" - but at least playing around with measurements is one way to become much more familiar with the technical aspects of photography and how to get around technical limitations.....

tony

TyKo Contributing Member • Posts: 611
Re: D100 Characteristic Curves Test

Would you care to say more about what you find necessary for good offset printing (either in terms of curves or post-processing)? Thanks.

-TyKo

Iliah Borg wrote:
Banal findings:
From the curves you created it is IMHO very definite that Nikon
goes with nearly 12%, and all the custom curves are applying TRC to
bring 12% to about 20-22%. Shifting custom curves to the right 0.5
eV brings them to pretty good match to the "normal" curve, exept
for the shoulder, which results in preserving highlights due to
extra room. It is also pretty obvious that trying to keep lower
noise levels, "normal" curve renders shadows darker then they
should be. And thus the smooth curve renders higher gamma in
midtones and highlights, compressing them. Shoulder of custom
curves come very close to Gretag-Macbeth recommended rendition of
highlights, having RGB=249 up to D=.23, which regular gamma 2.2
curve would render to approx. 200 This demands for a sharper
contrast in midtones, rendering them more punchy. From the point of
view of printing and tone separation "normal" curve is not very
good. For photoprinting custom curves seem to provide more pleasant
images. For offset printing... all the tone curves need extensive
post-processing.

I hope I understood your question below.

Would you be able to compare the tonality of the D100 with that of
Velvia or Provia using the Excel spreadsheet I created?

Iliah Borg Forum Pro • Posts: 29,605
Re: D100 Characteristic Curves Test

Would you care to say more about what you find necessary for good
offset printing (either in terms of curves or post-processing)?

I'm not quite sure custom curves can be bended that much to follow the reproduction curve. IMHO many bends, especially sharp ones affect de-mosaicking in a most terrible manner. The image after de-mosaicking with such curves exhibits artifacts. The curves I use are preserving only about 6.5 stops of dynamic range, as I believe the rest of shadows is just noise. I map the image to 30...242 in RGB, exept for specular highlights. To do so, I use the reproduction values from Kodak Q-13 scale, which was established since long ago as a standard for density/exposure control while taking photos for offset printing.

My steps to cut in post-processing for offset may seem to be a pain:

  • determine native color temperature of the sensor; all the following is done at this temperature, using colour correction filters;

  • determine custom curve to render image well, without artifacts (which means that the curve should match the de-mosaicking algorithm and preserve enough detail in highlights without agressive midtone boost) and loading this curve into the camera;

  • determine TRC according to Q-13 for images exposed right, as well as after + -eV correction in Capture;

  • taking normally exposed image of colour target with custom curve in-camera, putting it through TRC; and creating colour profile.

-- hide signature --

no text

Tom Judd Regular Member • Posts: 340
Re: For Iliah Borg - D100 Characteristic Curves Test

You guys are freaking me out. Can I still take good images without knowing how to do the math? Yes, I think I can. Sometimes to much is put into something when it really dose not matter for the average guy, like me. ;> ). Thanks for all the hard work photogenic. I do use your curves, because I like the result, I will leave the rocket science to someone else.

Tom Judd
pBase supporter
http://www.pbase.com/a3guy
I love this Forum!!!

Iliah Borg Forum Pro • Posts: 29,605
Velvia

I used Macbeth Color Checker.

The densities of gray scale and respective RGB values from the scan that went through offset well were:
=D= RGB
0.05 245
0.23 180
0.44 127
0.7 85
1.05 40
1.5 5

-- hide signature --

no text

OP fotogenetic Senior Member • Posts: 2,958
Re: For Iliah Borg - D100 Characteristic Curves Test

It helps to know he dynamic range of the sensor so you know how much EV compensation to use when metering for highlights. For example, with the results of my tests, I know that if I want to spot meter off the feathers of a swan in bright sunlight without blowing highlights, I can spot meter with +2.0 or +2.5 EV compensation and get what I want. It just so happens that there are a lot of people who have problems in the same type of situation because they do not know the characteristics of their sensor.

Knowledge can only help!

Tom Judd wrote:

You guys are freaking me out. Can I still take good images without
knowing how to do the math? Yes, I think I can. Sometimes to much
is put into something when it really dose not matter for the
average guy, like me. ;> ). Thanks for all the hard work
photogenic. I do use your curves, because I like the result, I will
leave the rocket science to someone else.

Tom Judd
pBase supporter
http://www.pbase.com/a3guy
I love this Forum!!!

jmills74 Contributing Member • Posts: 816
Re: Working link w/ graphs [imgs]

how about an actual test images????

I think this would really help show what curves are all about. If people see a real image with various curves applied, they might get more excited!!!

http://www.deviantart.com
http://jmills74.deviantart.com/

Tom Judd Regular Member • Posts: 340
Re: For Iliah Borg - D100 Characteristic Curves Test

Tom wrote:

Point taken. I guess my lack of a good enough math background to really understand how you do your calculations tends to get in my way. Since you put it in a way, that I can understand the reasoning of why you are doing this, it is much clearer. If it wasn't for you mathamaticians were would we be?
Thanks photogenic and all you geniuses out there.

Tom Judd
pBase supporter
http://www.pbase.com/a3guy
I love this Forum!!!

fotogenetic wrote:
It helps to know he dynamic range of the sensor so you know how
much EV compensation to use when metering for highlights. For
example, with the results of my tests, I know that if I want to
spot meter off the feathers of a swan in bright sunlight without
blowing highlights, I can spot meter with +2.0 or +2.5 EV
compensation and get what I want. It just so happens that there
are a lot of people who have problems in the same type of situation
because they do not know the characteristics of their sensor.

Knowledge can only help!

Tom Judd wrote:

You guys are freaking me out. Can I still take good images without
knowing how to do the math? Yes, I think I can. Sometimes to much
is put into something when it really dose not matter for the
average guy, like me. ;> ). Thanks for all the hard work
photogenic. I do use your curves, because I like the result, I will
leave the rocket science to someone else.

Tom Judd
pBase supporter
http://www.pbase.com/a3guy
I love this Forum!!!

Keyboard shortcuts:
FForum MMy threads