By the numbers ?


Dan Margulis <76270.1033@...>
 

Bob writes,

hushhhhhh! don't say that too loudly. I'm already losing a pile of bread
and butter business to clients that have taken a greater portion of their
basic photo work in-house because of what these things can do. These
cameras really are quite amazing in their capabilities and more and more
people are starting to figure that out. They still hire me for those jobs
where my special skills bring something unique to the image, but more and
more of the really simple (and very profitable) work is being taken
in-house.>>

This particular genie is not going back into the bottle. The negative
impact on professional photographers is already large and it's going to get
worse. Those head-in-the-sand folk who say, "I'm a photographer, what I do
is shoot pictures, let somebody else worry about what happens next" have
had a rough few years already, and now they're really going to pay the
price for their lack of foresight. It's the people like yourself who have
made a serious effort to adjust to the technology and be more full-service
who have a good chance to survive.

The impact as I see it is not merely the loss of certain jobs but that
prices are being driven down for what remains. I assume that the culprits
for this phenomenon are those desperados who have been forced to bid for
the work at any price by the shrinking demand for their services.

I'm planning to write a column about this in April, which is the 5th
anniversary of a similar column I wrote warning of this very thing, which
was much pooh-poohed at the time by professional photographers.

Dan Margulis


Rob Outlaw <routlaw@...>
 

Dan writes:

The problem seems to be not a defect in the camera itself, but rather that
in a lot of these cameras, and Nikon products in particular, there is some
very hamhanded, clumsy, and, of course, undocumented, automated color
correction going on before we even open the file. In the ones I've looked
at, there seems to be no way to defeat this automated "correction."
This seems like a case for implementing digital camera profiles after all,
be it from
Profile Maker or the Praxisoft solution. FWIW in a direct comparison between
the D1 and Phase One Lightphase this summer in my studio, I found the
LightPhase
captures without utilizing the canned profiles from Phase One to be
extremely dead
with some fairly drab color. In fact some colors on a simple Kodak color
chart did
not even show up regardless of how I captured and what profile that used.
Trying to make
something out of the Lightphase raw captures (files that had not been
converted with their
canned profiles) was way more trouble than it was worth in my experience.
Even after implementing the profiles from Phase One I found that I had to
edit the images
considerably more than with a D1 image.

Don't misunderstand me, I am not saying that you can not get spectacular
images from
the more expensive cameras, its just that it is not always the cake walk
many of these
manufacturers would have you believe. Its sort of like we are still in the
Wild West Days
of digital imaging and I think this makes a strong case for Color Management
in order to obtain
some sort of standard so that we are all playing on level field. At least
for me thats the
promise that I see with CM, though it may not be there 100% yet.

Selecting is usually unnecessary. For this category of images ONLY, I find
that a preliminary use of Hue/Saturation in RGB can help, because it can
specify a particular shade and move that. In this case you would click on
the fleshtone and move it in the yellow Hue direction, plus possibly
desaturate it. Another alternative when the skintone is too magenta, again
in RGB, is to blend the red channel into the green, Lighten mode, at
around
20% opacity.
In addition I have also found that shooting with the D1 with the low
contrast
setting (and then adjusting for contrast later in PS) yields a better image
with
less flesh tone problems.

The problem is, the preliminary "correction" these cameras make isn't
lighting-specific, it's image-specific
Although a number of users tend to have more problems while using the SB28dx
flash unit with the camera. A UV filter placed over the flash will help here
but
not totally eliminate the problem.

My own experience is that Apple
RGB or sRGB is a better choice most of the time, but it really depends on
the character of the image.
I still will stick with Adobe RGB space since it has given me little
problems even with the D1.

Rob Outlaw


Andrew Rodney <andrew@...>
 

on 1/1/01 11:26 PM, jbmmmac@... at jbmmmac@... wrote:

Problem of bad fleshtones is called Magenta problem but in reality it is
problem of Reds having too much Magenta and Cyan and not enough Yellow (and
Yellows  being contaminated with Cyan). The first step to correct this
problem is to pull red curve to reduce Cyan in midtones and next puling Green
curve to reduce magenta. Combined effect of pulling red and green curves
increase yellow and reduce cyan in skin tones, reds and yellows. Converting
from NTSC to ColorMatch is helpful to bring other colors and reducing overall
dullness due to different gamma of those color spaces.

That’s an awful lot of work to basically make a file tagged incorrectly LOOK correct in a  newly assigned Working Space. By simply tagging the D1 file with the correct description of the RGB you are getting would solve all these problems without altering the data in the file! Then you would convert into any RGB Working Space you want (the preview wouldn’t change at this point either). What you are basically doing is viewing a file in a condition where it’s not producing a proper preview. Then after you get it looking as you like, you Assign your Working Space. This is very similar to working with color neg scans. Since we can’t produce a profile for color negs, the best we can do is make the image look good on a calibrated display and then convert into the Working Space (usually using the custom display profile as the source, something you could try with the D1 if you didn’t want to deal with profiling the beast).

Andrew Rodney


Andrew Rodney <andrew@...>
 

on 1/1/01 10:50 PM, Dan Margulis at 76270.1033@... wrote:

I have played with a D1 but don't claim any expertise in it. What I can say
for sure is, if the problem is that fleshtones are too magenta, then
assuming that the files are NTSC is going to make that problem worse, not
better.

Nikon claims the D1 shots into NTSC but I don’t know if Nikon really has a clue either way. I can tell you with a custom input profile for the D1, the problem skin tone issues disappear. And not just the magenta but a problem with highlights blocking up (or I should say appearing to block up). Bringing a file into Photoshop assuming it’s one RGB space when it’s a mile away from that space makes a profound impact on what we see and what we eventually get when converting to an output space. Simply assigning the correct profile in Photoshop 6 makes this all disappear (because in reality, it’s not really there). The files appear magenta and print poorly because we are making a very bad assumption about the RGB we have. It’s as if we had a file in SWOP Uncoated newsprint but viewed it as if it were SWOP coated glossy stock. The preview would be wrong, the file would output wrong with that incorrect tag to SWOP coated. Take the file and output on the right device (or Assign the correct profile) and the preview and output are fine and dandy!

Andrew Rodney


Dan Margulis <76270.1033@...>
 

Bob writes,

Its not a matter of being wider. Its a matter of matching what Nikon was
targeting their output to. Why, I don't know... but Nikon has said that
Nikon D1 in camera processing is done with output to NTSC rgb in mind. It
has to be going to some definition of RGB and that's what Nikon chose.>>

I have played with a D1 but don't claim any expertise in it. What I can say
for sure is, if the problem is that fleshtones are too magenta, then
assuming that the files are NTSC is going to make that problem worse, not
better.

But it really depends on the image. I've shot over ten thousand exposures
with lower-end Nikons and it's pretty clear that NTSC isn't the right RGB
for them, although it may be for the D1 for all I know. OTOH, I was in the
hardest-hit county of the northeast storm this Saturday and went out and
shot in it when the snow was at its heaviest. Those exposures would have
been better opened in NTSC. But the bottom line is that the quality of the
data was phenomenal and if I had to produce a printable picture of the
blizzard I'd prefer to start with that cheap capture than with a chrome and
a drum scan.

What's needed on the lower end cameras is an option to deliver raw files
for
those that want to milk these things for all they can do. Access to the raw
files allows a tremendous amount of control over what the image looks
like.>>

Agreed.

Dan Margulis


Bob Smith <rmsmith@...>
 

Dan Margulis wrote:

Similarly, Bob's suggestion of using a wider-gamut RGB definition works
some of the time.
Its not a matter of being wider. Its a matter of matching what Nikon was
targeting their output to. Why, I don't know... but Nikon has said that
Nikon D1 in camera processing is done with output to NTSC rgb in mind. It
has to be going to some definition of RGB and that's what Nikon chose.
Kodak Pro cameras output RGB (via std acquire module processing) that very
closely approximates ColorMatch RGB. Pull Kodak files straight into
ColorMatch RGB and they'll need only minor color tweaking. Pull them into
something like Adobe RGB without compensation and you've got a whole
different task ahead of you. Even when the camera is set to deliver a
finished file, there's no profile embedded in the camera file to describe
its RGB space. Why write redundant data to every image when disk space and
write speed are at a premium?

Processing in a camera like the D1 is quite different from the auto
processing so prevalent in the lower end models. You can put a camera like
the D1 or the pro Kodaks into a condition where they auto-processes color,
but that's unusual and certainly not generally recommended.

What's needed on the lower end cameras is an option to deliver raw files for
those that want to milk these things for all they can do. Access to the raw
files allows a tremendous amount of control over what the image looks like.
I stayed with Kodak cameras and passed on the D1 mainly because it offered
such poor support for handling raw data files. Kodaks (the pro level
models) are built around the concept of having the camera deliver raw data
and then processing later. Its been fun to watch D1 shooters discover what
they can do with raw data. I've seen more than a few who were ready to
chunk the camera over color woes. Then they got hold of good piece of
software for handling raw data and thought they had a new camera. Raw data
files are now available on the new Olympus E-10... a sub $2k model. I bet
we see it on more even lower end models in the not too distant future.

Bob Smith


Dan Margulis <76270.1033@...>
 

Joey writes (and Crissy similarly),

With the advent and popularity of digital cameras in newspaper
photojournalism a new beast has really reared its head.It seems that unlike
film that absorbs light consistently across the spectrum (at least in
theory) the digital CCDs on some digital cameras (specifically the Nikon
D1) the absorption seems to inconsistent (i.e. skin tones are absorbing
more magenta than say wood).>>

The problem seems to be not a defect in the camera itself, but rather that
in a lot of these cameras, and Nikon products in particular, there is some
very hamhanded, clumsy, and, of course, undocumented, automated color
correction going on before we even open the file. In the ones I've looked
at, there seems to be no way to defeat this automated "correction."

Much of the time, this "correction" does what it's supposed to. And for a
nonprofessional user who hopes to use the image without further correction,
it's probably a good idea. For us, however, it can be a real PITA.

In the Nikon case what is happening is that the capture software is forcing
a neutral black point to occur no matter what. The CCDs seem to be a lot
more sensitive to color in shadows than conventional film is. I will skip
the technical discussion but the bottom line is that when an image is
dominated by a single color in the shadow area this will often result in a
countercast in the midtones. If you are seeing skintones too magenta I'll
bet that a lot of these pictures are taken outdoors. Crissy, OTOH, sees
skintones too cyan, which suggest different backgrounds.

Trying not to over explain but often time I have, say a jacket, with an
appropriate CMYK value and faces with way too much magenta. Short of
selecting each object with the wrong value is there another way, as I try
to
use the magic wand as little as possible?>>

Selecting is usually unnecessary. For this category of images ONLY, I find
that a preliminary use of Hue/Saturation in RGB can help, because it can
specify a particular shade and move that. In this case you would click on
the fleshtone and move it in the yellow Hue direction, plus possibly
desaturate it. Another alternative when the skintone is too magenta, again
in RGB, is to blend the red channel into the green, Lighten mode, at around
20% opacity.

Warning: don't try this with images shot conventionally. Using Hue/Sat
prematurely or channel blending to change color is ordinarily a good way to
mess things up beyond repair. It's only when you're trying to fix things
that have already been messed up by somebody else that the approach makes
sense.

The problem is, the preliminary "correction" these cameras make isn't
lighting-specific, it's image-specific, so doing something automated like
writing a curves script or trying to profile the camera, as Chris suggests
will work well with some images and make things worse with others.

Similarly, Bob's suggestion of using a wider-gamut RGB definition works
some of the time. It has the impact of making the image darker and more
vivid. That isn't likely to be right where the fleshtones are too magenta,
but it might be right if they are too cyan. My own experience is that Apple
RGB or sRGB is a better choice most of the time, but it really depends on
the character of the image.

I think that we're going to be hearing a lot more about this issue in the
future.

Dan Margulis


jbmmmac@...
 

In a message dated 1/1/01 11:55:01 PM, 76270.1033@compuserve.com writes:

<< I have played with a D1 but don't claim any expertise in it. What I can say
for sure is, if the problem is that fleshtones are too magenta, then
assuming that the files are NTSC is going to make that problem worse, not
better. >>

Problem of bad fleshtones is called Magenta problem but in reality it is
problem of Reds having too much Magenta and Cyan and not enough Yellow (and
Yellows being contaminated with Cyan). The first step to correct this
problem is to pull red curve to reduce Cyan in midtones and next puling Green
curve to reduce magenta. Combined effect of pulling red and green curves
increase yellow and reduce cyan in skin tones, reds and yellows. Converting
from NTSC to ColorMatch is helpful to bring other colors and reducing overall
dullness due to different gamma of those color spaces. Whenever I tell D1
user he needs to make D1 image redder to make it better he thinks I am joking
but I really don't. It is too much CYAN and MAGENTA in flesh tones that make
it ugly and the only way to reduce Cyan is to make image temporary more red
untill puling green will reduce magenta and increase yellow.

Janusz


jbmmmac@...
 

I know the problem you are talking about as I use D1 to shoot weddings and
"Magenta" problem is well known problem particulary to those shooting people
pictures. Part of bad skin tones color problem is due to making color
corections in wrong color space as in camera color processing use color space
close to NTSC characteristics and conversion to CMYK is done from let say
AdobeRGB/ColorMatch color spaces. Another problem with D1 files is that CCD
sensor has big sensitivity to Ultraviolet and Infrared light what makes
properly exposed files contain too much cyan in reds and yellows. If you add
not accurate White Ballance settings on some captures (auto WB is very bad so
try to set it manualy). As a experiment convert raw D1 file to CMYK from NTSC
color space and open the same file into your working space (without asigning
NTSC) and change mode to CMYK and you will see big difference beetween those
two CMYK files. The best way to deal with Nikon D1 files is to process them
by using Bibble or MacBibble software that was created for processing RAW Nef
D1 files (without need for expensive Nikon Capture software) but works great
with in camera processed JPGs or TIFs as well. For those processing in
Photoshop asigning NTSC color space and converting to working color space
plus pulling green and red curves can also do the trick. If anyone is using
ColorMacth I have to add that adjusting gamma is also neccesary as properly
exposed D1 file has gamma closer to 2.2 than native ColorMatch gamma 1.8
Properly adjusted D1 files can have beautiful color if you want to see sample
open this image below:
http://janusz.11net.com/D1/001a.jpg
Happy New Year
Janusz