Beating a dead horse: Adobe RGB for D60

hyslopc1 · Dec 19, 2002

This hasn't been discussed here much lately (where's Ron Parr?), which is interesting since it used to come up every day or two until a few months ago. Everytime it came up, the conclusions were:

1) You don't really need it if you shoot RAW and edit in 16-bit color, but it would be nice to have

2) The best way to get this working would be to design a color profile to convert from the D60 linear colorspace to the Adobe colorspace

3) Lots of people posted to say they're working on this and hope to have something ready sometime soon

So my question - have any of the people who were working on this gotten anywhere? And if no, did anyone conclude that this really is impossible? Chuck Westfall claimed that you would have to profile every camera individually to do this some time back, but not many people buy that theory.

MikeA · Dec 19, 2002

hyslopc1 said:
Chuck Westfall claimed that you would have to
profile every camera individually to do this some time back, but
not many people buy that theory.

True enough...we all know that all of these Canon D30 and D60 sensors are identical, and Westfall is only one of their top tech service guys in the U.S. ... what the hell would HE know, anyway? And these studio photographers who profile their specific lighting setups ... what a bunch of fools, when they could download a free "canned" profile off the web, huh? Honestly, some people.

Magne Nilsen · Dec 19, 2002

MikeA said:
hyslopc1 said:

Chuck Westfall claimed that you would have to
profile every camera individually to do this some time back, but
not many people buy that theory.

Click to expand...

True enough...we all know that all of these Canon D30 and D60
sensors are identical,

I can't speak for the D30, but the D60s are actually very identical.

MikeA said:
and Westfall is only one of their top tech
service guys in the U.S. ... what the hell would HE know, anyway?

Westfall knows a lot, but his knowledge in Color Management is based on what he has been told and has read, not so much from real life experience. On this topic he was wrong, and at least in my communication with him on the topic he admitted that he would need to re-phrase his statements regarding this in the future.

MikeA said:
And these studio photographers who profile their specific lighting
setups ... what a bunch of fools, when they could download a free
"canned" profile off the web, huh? Honestly, some people.

If you really want perfection, one profile for each of your controlled and exactely reproducable specific lighting situation can be valuable. It will take you time/$ effort to get them made, but anyway.

The easier approach - that should work for 90% of the images that 90% of the D60-photograpers take - would be to have one canned profile that was based on a correct white balance situation.
The problem is - that canned profile is not available.

I am working on one, but had to stop for some months beacuse of other assignments. I am now back on that project, and hope to have a first public showing this coming weekend.

In my first attempt to describe the Canon converted images I too used a simple approach, and this resulted in the free ETC1, 2 and 4 profiles. With the forthcoming ETC-3 profile I am using a CLUT (color-look-up-table) based description of the device gamut, covering more than 65000 control values (RGB values, or actually Lab values, but anyway) This way I am able to much better describe all the major and minor twists and tweaks that Canon has applied in their converted images. I can in theory model the gamuts tetrahedron (The full 3D gamut). But this takes time, since Canon has ended up with many twists and tweaks in there, and because I am a bit of a perfectionist

A simple triangle is not good enough to cover this, but to examplify, look at this:

The ETC-3 gamut described here is not shown as a clear simple triangle at this luminance level, since this is not the case. The exact values jumps and bends, but are mostly within the two shown triangles. As you see this is somewhere between sRGB and Adobe-98, but not exactly any of them.

The only way to master this is with a CLUT based profile. And in the D60 case a lot of manual corrections must be applied to the look-up-tables.

--
Magne

hyslopc · Dec 19, 2002

Hey Magne - exciting to hear you're still working on this, and so close to having something ready. Will this new profile be suitable for linear work? I'd be interested to hear your (latest) thoughts on linear vs non-linear.

Are you in Norway? I'm in Stockholm.
--
D60, 16-35/2.8L, 28-135 IS, 550EX

Magne Nilsen · Dec 19, 2002

hyslopc said:
Hey Magne - exciting to hear you're still working on this, and so
close to having something ready. Will this new profile be suitable
for linear work? I'd be interested to hear your (latest) thoughts
on linear vs non-linear.

The 1st will be a non-linear profile for in-camera JPGs and Canon RIC-converted JPGs/TIFFs. I am working on a linear one too, but that will be later.

hyslopc said:
Are you in Norway? I'm in Stockholm.

Yepp - Trondheim.

--
Magne

Ron Parr · Dec 19, 2002

hyslopc1 said:
This hasn't been discussed here much lately (where's Ron Parr?),

Hi Carl. I'm around, but I've just been very, very busy with other things. I haven't had much time to devote to photography period, let alone raw conversion.

I've been very busy with some projects at work, one of which is an effort to get our robot to use its laser rangefinder to make a blueprint-like map of buildings. Here's an example:

hyslopc1 said:
So my question - have any of the people who were working on this
gotten anywhere? And if no, did anyone conclude that this really
is impossible? Chuck Westfall claimed that you would have to
profile every camera individually to do this some time back, but
not many people buy that theory.

I continue to doubt Chuck Westfall's explanation. Things may quiet down for me in late February, so I might revisit the problem then. Sigh...

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

hyslopc · Dec 19, 2002

Laser rangefinder, eh? Why not strap a D30 onto him ;-)? Yeah OK, that'd be a whole different set of problems.

--
D60, 16-35/2.8L, 28-135 IS, 550EX

MikeA · Dec 20, 2002

Magne Nilsen said:
I can't speak for the D30, but the D60s are actually very identical.

I can't speak for either, but I would have to ask on what empirical basis can such a claim be made? How many tests of how many D60s -- from how many different production batches (unless there is only one huge batch, world without end!) -- would be required to make such a claim, credibly?

Magne Nilsen said:
On this topic he was wrong, and at least in my
communication with him on the topic he admitted that he would need
to re-phrase his statements regarding this in the future.

Perhaps so, and yet how can it be stated flatly that all the sensors are identical? "All" is a lot of sensors...from whom might Chuck Westfall be getting his information? Canon's own engineers, perhaps? (Why would he attempt to get his information about this from someone else, if not his own people?) And if they are indeed his sources of information, should they not be considered reliable sources concerning their own sensor products?

Magne Nilsen said:
If you really want perfection, one profile for each of your
controlled and exactely reproducable specific lighting situation
can be valuable.

Which is, as I understand it, exactly how this work is done for "big-ticket" digital product (or other) photography. And it makes perfect sense to me. When there's that much money at stake, why screw around with possible inaccuracy? Just as the photographer's fee is typically a small part of a total advertising budget, surely properly profiling the equipment for a given (expensive) shot ends up being minor trouble and expense by comparison with the pay-off. (Whether or not the approach makes sense for photographers in other situations.)

Magne Nilsen said:
The easier approach - that should work for 90% of the images that
90% of the D60-photograpers take - would be to have one canned
profile that was based on a correct white balance situation.

Well, 90% is pretty decent. I'll await your further tests...

Magne Nilsen said:
I am working on one, but had to stop for some months beacuse of
other assignments. I am now back on that project, and hope to have
a first public showing this coming weekend.

I will be more than happy to be proven wrong about this, but since we are speaking here of mass-produced sensors not priced at the level of, say, Phase One equipment or the like, it certainly stands to reason that there are going to be variations among sensors of a given type!

Magne Nilsen · Dec 20, 2002

MikeA said:
Magne Nilsen said:

I can't speak for the D30, but the D60s are actually very identical.

Click to expand...

I can't speak for either, but I would have to ask on what empirical
basis can such a claim be made? How many tests of how many D60s --
from how many different production batches (unless there is only
one huge batch, world without end!) -- would be required to make
such a claim, credibly?

I don't know how many you feel should be needed to make the claim credibly, but I have gathered images from 31 different cameras covering 14 production batches sold in the US, Canada, Brazil, Norway, Sweden, the UK, the Netherlands, Germany, Italy, Hong Kong, Japan and Australia. 12 of these cameras I've had in a controlled environment and profiled myself, 5 of the others I feel I have enough knowledge of to be quite sure of their capabilities. I have a group of beta-testers around the world, some of them 'frequent flyers' in this forum who also feel that the 'one canned profile' approch is working very good for the D60.

MikeA said:
MikeA said:

On this topic he was wrong, and at least in my
communication with him on the topic he admitted that he would need
to re-phrase his statements regarding this in the future.

Click to expand...

Perhaps so, and yet how can it be stated flatly that all the
sensors are identical?

I don't think anyone 'stated flatly' anything.

MikeA said:
"All" is a lot of sensors...from whom might
Chuck Westfall be getting his information? Canon's own engineers,
perhaps? (Why would he attempt to get his information about this
from someone else, if not his own people?) And if they are indeed
his sources of information, should they not be considered reliable
sources concerning their own sensor products?

I guess a pointer back to the thread where this came up first is useful: http://forums.dpreview.com/forums/read.asp?forum=1019&message=3127234

MikeA said:
MikeA said:

If you really want perfection, one profile for each of your
controlled and exactely reproducable specific lighting situation
can be valuable.

Click to expand...

Which is, as I understand it, exactly how this work is done for
"big-ticket" digital product (or other) photography. And it makes
perfect sense to me. When there's that much money at stake, why
screw around with possible inaccuracy? Just as the photographer's
fee is typically a small part of a total advertising budget, surely
properly profiling the equipment for a given (expensive) shot ends
up being minor trouble and expense by comparison with the pay-off.
(Whether or not the approach makes sense for photographers in other
situations.)

You seem to think a 'perfect' profile is something that is done in a few hours with the right amount of equipment and $'s ? Well, it's not.

MikeA said:
MikeA said:

The easier approach - that should work for 90% of the images that
90% of the D60-photograpers take - would be to have one canned
profile that was based on a correct white balance situation.

Click to expand...

Well, 90% is pretty decent. I'll await your further tests...

MikeA said:

I am working on one, but had to stop for some months beacuse of
other assignments. I am now back on that project, and hope to have
a first public showing this coming weekend.

Click to expand...

I will be more than happy to be proven wrong about this, but since
we are speaking here of mass-produced sensors not priced at the
level of, say, Phase One equipment or the like, it certainly stands
to reason that there are going to be variations among sensors of a
given type!

They are not 'mass-produced sensors' in the Olympus / Casio / Sony / G2 / Coolpix sense, but why should they fluctate as much as you seem to think? They are made with the same technology as any modern microprocessor, at least my Intel/AMD chips are exceptionally equal, even across batches.

The differences we see are mostly in the firmware/software implementation each maker applies, and even that doesn't change among camera/sensor batches, i.e. the D60 RAW converter is still the same as for the first cameras made, and it still does the same job for each and every D60 produced.

Here's a sample - the first straight out of camera, the other with ETC3 applied, and then converted to sRGB.

1: original from camera, just shrinked.

2: ETC3 profile applied, and converted to sRGB

Do you see any differences? Send my a image if you want more...
--
Magne

MikeA · Dec 20, 2002

Magne Nilsen said:
I don't know how many you feel should be needed to make the claim
credibly, but I have gathered images from 31 different cameras
covering 14 production batches sold in the US, Canada, Brazil,
Norway, Sweden, the UK, the Netherlands, Germany, Italy, Hong Kong,
Japan and Australia. 12 of these cameras I've had in a controlled
environment and profiled myself, 5 of the others I feel I have
enough knowledge of to be quite sure of their capabilities.

Then, that certainly weighs in favor of your claim. Until I saw the above I had no idea what testing you'd done or hadn't done.

Magne Nilsen said:
You seem to think a 'perfect' profile is something that is done in
a few hours with the right amount of equipment and $'s ? Well, it's
not.

"Perfect" is your word here -- not mine. I stated what some professional photographers do during big-ticket studio shots, and why they do it. Whether they do it "perfectly" is another matter and I can't (and didn't) comment on that.

Magne Nilsen said:
Here's a sample - the first straight out of camera, the other with
ETC3 applied, and then converted to sRGB.
2: ETC3 profile applied, and converted to sRGB

Magne Nilsen said:
Do you see any differences? Send my a image if you want more...

The monitor on which I'm viewing the images right this moment is far from accurately color-corrected, so I will have to wait until I get to a different monitor. That said, I do see some slight differences in the two images. No telling how they'll appear on the better monitor later on...

MikeA · Dec 20, 2002

I said:
get to a different monitor. That said, I do see some slight
differences in the two images. No telling how they'll appear on the
better monitor later on...

To clarify: that one image is profiled and the other isn't, but that I see only slight differences in the two images, suggests to me that the monitor is way, way off...

Magne Nilsen · Dec 20, 2002

MikeA said:
I said:

get to a different monitor. That said, I do see some slight
differences in the two images. No telling how they'll appear on the
better monitor later on...

Click to expand...

To clarify: that one image is profiled and the other isn't, but
that I see only slight differences in the two images, suggests to
me that the monitor is way, way off...

If you don't see a big change in the greens, your monitor has a problem, yes.

Anyway, try this: Save both images to local disk. Read 1st (org.) image into PhotoShop, assign sRGB and convert it to sRGB. Read 2nd (etc3) image into PS, it is already converted to sRGB, so use the embedded profile.

Here are some of the measurable differences on the Kodak chart:

White, grays and black are ~ unchanged
Blue: R31 G32 B103 -> R30 G28 B100
Cyan: R32 G143 B203 -> R0 G129 B174
Green: R23 G134 B72 -> R0 G120 B44
Yel.: R198 G184 B50 -> R219 G191 B22
Red : R167 G24 B14 -> R178 G19 B19
Mag.: R169 G25 B77 -> R179 G7 B97

Fuji box green: R15 G155 B78 -> R0 G140 B40
Canon red: R162 G8 B6 -> R176 G0 B9

These are major changes by the numbers, but visually not that easy to spot on this kind of image, since there are so many pure colors covering large areas. I have a good profiled monitor and a controlled 5300 Kelvin light source and those very articles in view as of this writing. Are they perfectly identical to the profiled image? Probably not perfectly - but I'ld say > 90% identical. I.e. the Fuji film box and the Kodak control patches looks exactely like the true objects to me.

I used this image since I hoped some of those objects would be familiar to most people.

--
Magne

Rune Moberg · Dec 22, 2002

Magne Nilsen said:
http://genji.image.pbase.com/u20/magne/upload/9217138.939_smalletc3srgb.jpg

I'm not much of a Cola drinker, but on the label where there's a drawing of the cork popping open, there's a highlight which in the etc-3 profiled version has a slightly yellowish tint to it. Am I seeing things or is it supposed to be like that?

--
Rune, http://runesbike.com/

Magne Nilsen · Dec 22, 2002

Rune Moberg said:
Magne Nilsen said:

http://genji.image.pbase.com/u20/magne/upload/9217138.939_smalletc3srgb.jpg

Click to expand...

I'm not much of a Cola drinker, but on the label where there's a
drawing of the cork popping open, there's a highlight which in the
etc-3 profiled version has a slightly yellowish tint to it. Am I
seeing things or is it supposed to be like that?

Hmm, I did upload a 200K JPG to pbase. But that file is now just 40K ??
I think you see JPG artifacts.
Try this one:

--
Magne

Magne Nilsen · Dec 23, 2002

I upload 200K JPGs (PS quality 12) to pbase, but some of them get shrinked to 40K with a lot of JPG compression artifacts?

I will delete them from pbase, and move them to another site.

--
Magne

Rune Moberg · Dec 23, 2002

Magne Nilsen said:
I think you see JPG artifacts.

I think my eyes deceive me. The yellow label on the gin bottle fools me into seeing yellow anywhere. Sorry.

--
Rune, http://runesbike.com/

Search

Beating a dead horse: Adobe RGB for D60

hyslopc1

Active member

MikeA

Senior Member

Magne Nilsen

Senior Member

hyslopc

Senior Member

Magne Nilsen

Senior Member

Ron Parr

Guest

hyslopc

Senior Member

MikeA

Senior Member

Magne Nilsen

Senior Member

MikeA

Senior Member

MikeA

Senior Member

Magne Nilsen

Senior Member

Rune Moberg

Senior Member

Magne Nilsen

Senior Member

Magne Nilsen

Senior Member

Rune Moberg

Senior Member

Keyboard shortcuts