ColorChecker SG
31 views
Skip to first unread message
Antonio
Jun 14, 2009, 12:53:17 PM6/14/09
to AcrCalibrator
I have the 24 patch ColorChecker and the 140 patch Colorchecker SG
I have done a successful calibration using AcrCalibrator and the 24
patch ColorChecker.
Will I get a substantially better camera profile/calibration if I use
the Colorchecker SG and the EyeOne Match software?
Thanks,
Antonio
I have done a successful calibration using AcrCalibrator and the 24
patch ColorChecker.
Will I get a substantially better camera profile/calibration if I use
the Colorchecker SG and the EyeOne Match software?
Thanks,
Antonio
steve sprengel
Jun 14, 2009, 2:46:55 PM6/14/09
to acrcal...@googlegroups.com
The ACR Calibrator script only uses 3 of the color-patches (red, green, blue) to do its calibration so most things would do a better job; however, since the script is merely tweaking an existing color-profile, it may not be that bad.
Adobe Labs has released the free DNG Profile Editor that uses all 24 squares to create a new camera profile (rather than tweaking an existing profile), and I use this, now, not the script. You can create a single-color-table profile that works for a particular lighting situation, as well as the dual-color-table profiles that work for most lighting situations because ACR interpolates between the two color-tables in the profile the same as they do with the stock color profiles.
I don't have any experience with the EyeOne Match software to know what it does and how well it works, but if it uses many color patches, it should do a reasonable job. If it doesn't create a dual-color profile then it may only be useful for single-color-lighting situations, rather than a more universal application like Adobe's dual-color profiles. For studio work a single-color profile would be
I have created color-error plots showing the errors for each of the 24 color-patches for both the script and the DNG Profile Editor for various lighting situations using a program called Imatest which you can view, here:
The thing to see from such color-error plots is that the color-error is not the same across all colors with some colors being almost perfect and some having quite a bit of error. In professional work, for portraits you want the skin-tones to look good, and in catalog work, you want the textile colors to be correct, so the fact that all colors aren't quite right is not big a concern, but in general it would be nice to have the colors close. This optimization of pariticular colors at the expense of others is what the calibration sliders are for in the first place and is actually the main purpose of the DNG Profile Editor--to tweak a profile so particular profiles are perfect.
With the newer ACR versions, such as you have with CS4, Adobe has released several different color-profiles for each Canon and Nikon camera model they support to make different situations look good. For example, the portrait profile makes the skintones look good, and the landscape profile oversaturates the greens and the blues to make sky and foliage look good. These are called camera-match profiles because they try to mimick the typical scene-modes that various camera manufacturers supply for the in-camera JPG conversion. At the point that Labs released the DNG Profile Editor, these camera-match profiles were in beta and you could download and install them from Labs, but now ACR comes with these profiles and you don't want to install the betas which would just yield duplicates.
Sharam
Jun 14, 2009, 10:01:35 PM6/14/09
to AcrCalibrator
Steve,
Does Imatest assume the ideal lighting is D50 (is this the same as
5000K?). Does this contribute to the error when comparing an image
taken in 6500K?
Do you know how to create a "perfect" 6500K DNG image of the 24 Color
Checker? I would like to run the DNG Profile Editor on this image and
check to see if DNG Profile Editor gives zero adjustments. If it does
yield zero adjustment it implies that it will not adjust a perfectly
good image.
Thanks,
Sharam
On Jun 14, 11:46 am, steve sprengel <steve.spren...@gmail.com> wrote:
> The ACR Calibrator script only uses 3 of the color-patches (red, green,
> blue) to do its calibration so most things would do a better job; however,
> since the script is merely tweaking an existing color-profile, it may not be
> that bad.
>
> Adobe Labs has released the free DNG Profile Editor that uses all 24 squares
> to create a new camera profile (rather than tweaking an existing profile),
> and I use this, now, not the script. You can create a single-color-table
> profile that works for a particular lighting situation, as well as the
> dual-color-table profiles that work for most lighting situations because ACR
> interpolates between the two color-tables in the profile the same as they do
> with the stock color profiles.
>
> http://labs.adobe.com/wiki/index.php/DNG_Profileshttp://labs.adobe.com/wiki/index.php/DNG_Profiles:Editor
> > Antonio- Hide quoted text -
>
> - Show quoted text -
Does Imatest assume the ideal lighting is D50 (is this the same as
5000K?). Does this contribute to the error when comparing an image
taken in 6500K?
Do you know how to create a "perfect" 6500K DNG image of the 24 Color
Checker? I would like to run the DNG Profile Editor on this image and
check to see if DNG Profile Editor gives zero adjustments. If it does
yield zero adjustment it implies that it will not adjust a perfectly
good image.
Thanks,
Sharam
On Jun 14, 11:46 am, steve sprengel <steve.spren...@gmail.com> wrote:
> The ACR Calibrator script only uses 3 of the color-patches (red, green,
> blue) to do its calibration so most things would do a better job; however,
> since the script is merely tweaking an existing color-profile, it may not be
> that bad.
>
> Adobe Labs has released the free DNG Profile Editor that uses all 24 squares
> to create a new camera profile (rather than tweaking an existing profile),
> and I use this, now, not the script. You can create a single-color-table
> profile that works for a particular lighting situation, as well as the
> dual-color-table profiles that work for most lighting situations because ACR
> interpolates between the two color-tables in the profile the same as they do
> with the stock color profiles.
>
>
> - Show quoted text -
Steve Sprengel
Jun 15, 2009, 1:15:26 AM6/15/09
to acrcal...@googlegroups.com
The ProPhotoRGB standard numbers for a ColorChecker are for when it is
illuminated by D50 lighting. This does not mean that you can only test the
color-checker in D50 lighting. It is the job of the white-balance
adjustment in the raw-converter to adjust the colors to make the neutral's
neutral.
There are "ideal" color-checkers available as TIFs from BabelColor but I
think the DNG Profile Editor only works for RAWs from a camera that Adobe
has created a camera-profile for, so if you're not Adobe, I don't know how
you create a synthetic RAW to test with.
--------------------------------------------------
From: "Sharam" <sharam...@gmail.com>
Sent: Sunday, June 14, 2009 9:01 PM
To: "AcrCalibrator" <acrcal...@googlegroups.com>
Subject: [acrcal] Re: ColorChecker SG
illuminated by D50 lighting. This does not mean that you can only test the
color-checker in D50 lighting. It is the job of the white-balance
adjustment in the raw-converter to adjust the colors to make the neutral's
neutral.
There are "ideal" color-checkers available as TIFs from BabelColor but I
think the DNG Profile Editor only works for RAWs from a camera that Adobe
has created a camera-profile for, so if you're not Adobe, I don't know how
you create a synthetic RAW to test with.
--------------------------------------------------
From: "Sharam" <sharam...@gmail.com>
Sent: Sunday, June 14, 2009 9:01 PM
To: "AcrCalibrator" <acrcal...@googlegroups.com>
Subject: [acrcal] Re: ColorChecker SG
Antonio
Jun 16, 2009, 12:22:39 PM6/16/09
to AcrCalibrator
Hello Steve,
Thank you again for all your great information and help.
I will certainly try the Adobe DNG profile editor, it may well be the
answer I was looking for.
In my experience, I was always unhappy with the colors I was getting
on perfectly calibrated Eizo CG221 monitors.
I used EyeOne Match plus the ColorChecker SG to create a supposedly
great camera profile but was very disappointed, the colors were even
more off.
The profile created wiusing AcrCalibrator and the 24 patch
ColorChecker has given me much better results, but still I need
improvements. For example, blonde hair colors become too orange. Skin
tones a touch yellowish.
It is amazing how different from reality, under same 5500K light, some
of the colors of the 24 patch CC show in the calibrated monitor even
after profiling the camera.
Have you ever achieved a good consistent match across all colors of
the physical CC and its image on the screen?
Thanks,
Antonio
On Jun 14, 7:46 pm, steve sprengel <steve.spren...@gmail.com> wrote:
> The ACR Calibrator script only uses 3 of the color-patches (red, green,
> blue) to do its calibration so most things would do a better job; however,
> since the script is merely tweaking an existing color-profile, it may not be
> that bad.
>
> Adobe Labs has released the free DNG Profile Editor that uses all 24 squares
> to create a new camera profile (rather than tweaking an existing profile),
> and I use this, now, not the script. You can create a single-color-table
> profile that works for a particular lighting situation, as well as the
> dual-color-table profiles that work for most lighting situations because ACR
> interpolates between the two color-tables in the profile the same as they do
> with the stock color profiles.
>
> http://labs.adobe.com/wiki/index.php/DNG_Profileshttp://labs.adobe.com/wiki/index.php/DNG_Profiles:Editor
Thank you again for all your great information and help.
I will certainly try the Adobe DNG profile editor, it may well be the
answer I was looking for.
In my experience, I was always unhappy with the colors I was getting
on perfectly calibrated Eizo CG221 monitors.
I used EyeOne Match plus the ColorChecker SG to create a supposedly
great camera profile but was very disappointed, the colors were even
more off.
The profile created wiusing AcrCalibrator and the 24 patch
ColorChecker has given me much better results, but still I need
improvements. For example, blonde hair colors become too orange. Skin
tones a touch yellowish.
It is amazing how different from reality, under same 5500K light, some
of the colors of the 24 patch CC show in the calibrated monitor even
after profiling the camera.
Have you ever achieved a good consistent match across all colors of
the physical CC and its image on the screen?
Thanks,
Antonio
On Jun 14, 7:46 pm, steve sprengel <steve.spren...@gmail.com> wrote:
> The ACR Calibrator script only uses 3 of the color-patches (red, green,
> blue) to do its calibration so most things would do a better job; however,
> since the script is merely tweaking an existing color-profile, it may not be
> that bad.
>
> Adobe Labs has released the free DNG Profile Editor that uses all 24 squares
> to create a new camera profile (rather than tweaking an existing profile),
> and I use this, now, not the script. You can create a single-color-table
> profile that works for a particular lighting situation, as well as the
> dual-color-table profiles that work for most lighting situations because ACR
> interpolates between the two color-tables in the profile the same as they do
> with the stock color profiles.
>
Steve Sprengel
Jun 16, 2009, 2:06:44 PM6/16/09
to acrcal...@googlegroups.com
Remember that the ACR Calibrator, aside from using
the next-to-white patch for the white-balance, is only using 3 color patches,
the red, green, and blue of the 3rd row to compute its profile tweaks. So
the fact it is doing a better job seems to suggest that something is wrong,
somewhere, and it will be interesting to see if the DNG Profile Editor does any
better.
If you want me to check things in Imatest, post your original CC RAW as well as ACR-toning and camera-profile settings, including both the profile name (ACR x.x or Adobe Standard or Camera Standard or whatever) and calibration slider values the script produced,, and I can run it through Imatest just to see what color-error plot looks like.
If you don't have your own webspace, just upload the RAW to RapidShare.com and then post the Download link it gives you at the end. Without paying or registering with RapidShare, this link is good for 10-downloads or 90-days whichever comes first, and since there are so few of us on this board, it should be ok.
The DNG Profile Editor works instantly rather than the 10s of minutes the script takes to run...at least on my old computer.
The one thing to take care with when you create a dual-color-table profile is that the color-tables are specifically for 2850K and 6500K so you need to shoot CC in approximately those lighting conditions, which are different than the slightly warmer 5000K to 5500K you are probably working with. The incandescent (2850K) color-table shot was not too much of a problem, but I had to go outside on a semi-cloudy day to get something close to 6500K which needed particular weather.
If you are doing studio work with always the same color lights, then just make a single profile with both color-tables the same and use whatever color lighting you want, because there won't be any interpolation if both color-tables are the same.
The second DNG Profile Editor link I posted, originally, has a long tutorial on how to use the program, probably covering more situations than you want, so if you can't figure something out, you can ask, here, or even I the user-to-user forum link they list at the end of the main DNG Profile Editor page:
http://labs.adobe.com/wiki/index.php/DNG_Profiles:Editor
If you want me to check things in Imatest, post your original CC RAW as well as ACR-toning and camera-profile settings, including both the profile name (ACR x.x or Adobe Standard or Camera Standard or whatever) and calibration slider values the script produced,, and I can run it through Imatest just to see what color-error plot looks like.
If you don't have your own webspace, just upload the RAW to RapidShare.com and then post the Download link it gives you at the end. Without paying or registering with RapidShare, this link is good for 10-downloads or 90-days whichever comes first, and since there are so few of us on this board, it should be ok.
The DNG Profile Editor works instantly rather than the 10s of minutes the script takes to run...at least on my old computer.
The one thing to take care with when you create a dual-color-table profile is that the color-tables are specifically for 2850K and 6500K so you need to shoot CC in approximately those lighting conditions, which are different than the slightly warmer 5000K to 5500K you are probably working with. The incandescent (2850K) color-table shot was not too much of a problem, but I had to go outside on a semi-cloudy day to get something close to 6500K which needed particular weather.
If you are doing studio work with always the same color lights, then just make a single profile with both color-tables the same and use whatever color lighting you want, because there won't be any interpolation if both color-tables are the same.
The second DNG Profile Editor link I posted, originally, has a long tutorial on how to use the program, probably covering more situations than you want, so if you can't figure something out, you can ask, here, or even I the user-to-user forum link they list at the end of the main DNG Profile Editor page:
http://labs.adobe.com/wiki/index.php/DNG_Profiles:Editor
If you're interested in
getting numerical values for the color error, you can
also download a set of scripts that contains one that will give you the errors
with respect to standard CC24 values from Rags-Gardner's site:
The one you want to show the errors is: Read_Colors_CC24.jsx
but there are also others for various other calibration targets including the SG
and the IT8, so even if you've used a CC24 as your calibration target, you can
compute the error against a different target.
As far as some colors looking wrong compared to
others when viewing a CC photograph on screen, when I calibrate cheaper LCD
monitors, I usually see some of the mid-tone neutral patches go more red
and some go more green, which suggests the screen-profile is not that accurate
or smooth and perhaps the color-errors you're seeing in your CC patches is just
an indication of how few colors your monitor software used to calibrate with as
well as how few colors your monitor is using to create all the shades and hues
you see. I don't think a clibration is ever going to get all the colors
right, and you are either going to be trying to minimize the color-error
overall, or for specific colors that are important. I mean why are there
calibration sliders in ACR at all if calibrations could be made perfect,
easily.
Furthermore, some colors are almost impossible to
get right in a photograph, like the pale blue flower patch, and it's possible
that making those more correct is throwing off some of your other colors in the
opposite direction. That's why I like to see the color-error for each of
the patches so I can judge if the important colors are close or not. It
may be that Adobe's Camera Profile for you brand and model of camera that you're
tweaking is just really far off from what you'd like for how the colors look and
the script is only correcting some of the difference. Tindemans has a
variation of the script that you can specify the "weight" of each of the color
patches, so you can give a higher weighting to the skin-tones and let some of
the other colors change a bit. The fact that the CC24 only has a couple
skintone patches and many other colors means other-colors have more weight in
how things look unless you weight the skintones more.
--------------------------------------------------
From: "Antonio" <ans...@gmail.com>
Sent: Tuesday, June 16, 2009 11:22 AM
Antonio
Jun 17, 2009, 3:45:01 AM6/17/09
to AcrCalibrator
Hello Steve,
The original RAW file is
http://www.zshare.net/download/61481695df18d1ee/
That file converted to .DNG by the Hasselblad software is:
http://www.zshare.net/download/61481858959133d3/
The script results using the first file, which the ACR Release
Candidate can read, are:
-1, -13, +13, -30, +62, +12, -34 (First one, -1 is for Shadows Tint,
-13 for Red Primary Hue, etc.)
I ran the script after white balancing using the WhiteBal card on the
picture and reducing exposure by 0.5 stops to bring the white patch
(bottom left) to 245 aprox.
For the CC 140 the RAW file is
http://www.zshare.net/download/6148205014a654f9/
The TIFF file used by EyeOne Match from that RAW file is
http://www.zshare.net/download/61482168cc17e666/
The sidecar file for the white balance (with the whibal card) and
slight exposure adjustment to match the white and black patches
exposures is
http://www.zshare.net/download/61482191153c49f5/
The camera profile created by EyeOne Match with the TIFF file above is
http://www.zshare.net/download/614822550eb97347/
Any inputs or further questions will be much appreciated.
Thanks,
Antonio
> >>http://labs.adobe.com/wiki/index.php/DNG_Profileshttp://labs.adobe.co...
> >> - Show quoted text -- Hide quoted text -
The original RAW file is
http://www.zshare.net/download/61481695df18d1ee/
That file converted to .DNG by the Hasselblad software is:
http://www.zshare.net/download/61481858959133d3/
The script results using the first file, which the ACR Release
Candidate can read, are:
-1, -13, +13, -30, +62, +12, -34 (First one, -1 is for Shadows Tint,
-13 for Red Primary Hue, etc.)
I ran the script after white balancing using the WhiteBal card on the
picture and reducing exposure by 0.5 stops to bring the white patch
(bottom left) to 245 aprox.
For the CC 140 the RAW file is
http://www.zshare.net/download/6148205014a654f9/
The TIFF file used by EyeOne Match from that RAW file is
http://www.zshare.net/download/61482168cc17e666/
The sidecar file for the white balance (with the whibal card) and
slight exposure adjustment to match the white and black patches
exposures is
http://www.zshare.net/download/61482191153c49f5/
The camera profile created by EyeOne Match with the TIFF file above is
http://www.zshare.net/download/614822550eb97347/
Any inputs or further questions will be much appreciated.
Thanks,
Antonio
Steve Sprengel
Jun 17, 2009, 8:14:46 AM6/17/09
to acrcal...@googlegroups.com
The script computes its own white-balance and the toning values (exposure,
black-point, brightness, contrast) before it computes the calibration
sliders.
What numbers did it get for white-balance and toning? It is the top part of
the results window when the script has finished.
--------------------------------------------------
From: "Antonio" <ans...@gmail.com>
Sent: Wednesday, June 17, 2009 2:45 AM
black-point, brightness, contrast) before it computes the calibration
sliders.
What numbers did it get for white-balance and toning? It is the top part of
the results window when the script has finished.
--------------------------------------------------
From: "Antonio" <ans...@gmail.com>
Sent: Wednesday, June 17, 2009 2:45 AM
Antonio
Jun 17, 2009, 6:00:20 PM6/17/09
to AcrCalibrator
Sorry, please find all the results here
http://www.zshare.net/image/61511383780ce5d2/
> ...
>
> read more »- Hide quoted text -
http://www.zshare.net/image/61511383780ce5d2/
>
> read more »- Hide quoted text -
Steve Sprengel
Jun 18, 2009, 12:08:00 AM6/18/09
to acrcal...@googlegroups.com
The calibration sliders the script produced for you are suspiciously large.
To see if I get the same thing, I have started re-running the script on the
H-software DNG you provided but my computer is about 6 times slower than
yours so it will likely be the morning before I have something to compare to
yours.
To see if I get the same thing, I have started re-running the script on the
H-software DNG you provided but my computer is about 6 times slower than
yours so it will likely be the morning before I have something to compare to
yours.
I already tried running the script on the 3FR but Photoshop keeps getting an
error a few iterations into the run. I am using 32-bit XP on a single-core
machine with a bit more than 1 GB of memory available to Photoshop, but
perhaps the size of this RAW is just too large for whatever the script is
trying to do with it. I would guess a 64-bit version of PS wouldn't have
the same sort of problem.
Originally I applied the slider values you communicated to the .3FR file and
Imatest showed significant oversaturation for most of the colors, but now I
see that the slider values are from running on the H-software DNG, so I need
to re-do the Imatest but the computer is busy doing the script for the next
few hours so that will also be in tomorrow. It's possible that the .3FR
and the .DNG will have identical errors, but until I try, I don't know,
because I'm not sure what is different about the H-software version of the
DNG from an Adobe software version.
--
You should also reshoot the CC24 RAW so that is isn't overexposed. The DNG
Profile Editor reports that the three yellow squares (I think) and the white
square have at least one channel clipped to 255, so I have been unable to
use that tool to create a camera profile, either. If you reshoot the CC24
dimmer, post another link to it, along with the H-version of the DNG. I
almost always just shoot my target handheld in P mode, and don't have an
exposure problem. If you have a gray-card you can set the exposure in
M-mode using that. Green grass might be a substitute for a gray card to set
the exposure with, although if it is well-fertilized, dark grass you might
shoot at EV -0.5 to make sure it also doesn't overexpose.
Imatest also complained that there is a vertical gradient (darker or lighter
at the top that then bottom, not sure which), and maybe you can guess why
that is, based on the way it is tilted and what is behind you. You might
also try making it more of a 45-degree angle, pointed up a bit more at the
sky, and shoot more down on it behind careful not to create a shadow of any
kind on it. If I'm at a location is relatively clear of trees and
buildings except along the horizon, I actually put it flat on the ground,
sidewalk, driveway or street, then stand back a ways and zoom in quite a
bit, and shoot the target at an angle from 45 to 60, with the wider edge
about 3/4 of the way across the frame. Because you define the four corners
separately it doesn't matter that it isn't parallel with the camera. You
can think of what would be visible from the camera's point of view if the
target was a mirror, and make sure that mostly only sky would be in that
imagined reflection. Don't have the sun glaring off of it, too much,
either. You can judge this by if the black plastic between the squares has
too much glare. The easiest way to keep glare from happening is to not have
the sun directly "across" the target from your camera position.
--------------------------------------------------
From: "Antonio" <ans...@gmail.com>
Sent: Wednesday, June 17, 2009 5:00 PM
Antonio
Jun 18, 2009, 3:38:03 AM6/18/09
to AcrCalibrator
1. I ran the script on my laptop. My dekstop is several times faster
Of note is that the version of ACR is newer and not compatible with
the script, you get the warning, but it runs anyway
On Jun 18, 5:08 am, "Steve Sprengel" <st...@sprengels.com> wrote:
> The calibration sliders the script produced for you are suspiciously large.
> To see if I get the same thing, I have started re-running the script on the
> H-software DNG you provided but my computer is about 6 times slower than
> yours so it will likely be the morning before I have something to compare to
> yours.
>
> I already tried running the script on the 3FR but Photoshop keeps getting an
> error a few iterations into the run. I am using 32-bit XP on a single-core
> machine with a bit more than 1 GB of memory available to Photoshop, but
> perhaps the size of this RAW is just too large for whatever the script is
> trying to do with it. I would guess a 64-bit version of PS wouldn't have
> the same sort of problem.
>
> Originally I applied the slider values you communicated to the .3FR file and
> Imatest showed significant oversaturation for most of the colors, but now I
> see that the slider values are from running on the H-software DNG, so I need
> to re-do the Imatest but the computer is busy doing the script for the next
> few hours so that will also be in tomorrow. It's possible that the .3FR
> and the .DNG will have identical errors, but until I try, I don't know,
> because I'm not sure what is different about the H-software version of the
> DNG from an Adobe software version.
2. I agree fully, suspicious oversaturation is one evident result.
The .3FR and .DNG files are theoretically different.
> You should also reshoot the CC24 RAW so that is isn't overexposed. The DNG
> Profile Editor reports that the three yellow squares (I think) and the white
> square have at least one channel clipped to 255, so I have been unable to
> use that tool to create a camera profile, either. If you reshoot the CC24
> dimmer, post another link to it, along with the H-version of the DNG. I
> almost always just shoot my target handheld in P mode, and don't have an
> exposure problem. If you have a gray-card you can set the exposure in
> M-mode using that. Green grass might be a substitute for a gray card to set
> the exposure with, although if it is well-fertilized, dark grass you might
> shoot at EV -0.5 to make sure it also doesn't overexpose.
>
I did measure the exposure carefully with a Sekonic meter, bracketed
and used the best take. I used Manual mode with a tripod.
Even that, I agree, some patches were slightly overexposed. I lowered
the exposure in ACR and the readings were all fine.
Is this not good enough?
> Imatest also complained that there is a vertical gradient (darker or lighter
> at the top that then bottom, not sure which), and maybe you can guess why
> that is, based on the way it is tilted and what is behind you. You might
> also try making it more of a 45-degree angle, pointed up a bit more at the
> sky, and shoot more down on it behind careful not to create a shadow of any
> kind on it. If I'm at a location is relatively clear of trees and
> buildings except along the horizon, I actually put it flat on the ground,
> sidewalk, driveway or street, then stand back a ways and zoom in quite a
> bit, and shoot the target at an angle from 45 to 60, with the wider edge
> about 3/4 of the way across the frame. Because you define the four corners
> separately it doesn't matter that it isn't parallel with the camera. You
> can think of what would be visible from the camera's point of view if the
> target was a mirror, and make sure that mostly only sky would be in that
> imagined reflection. Don't have the sun glaring off of it, too much,
> either. You can judge this by if the black plastic between the squares has
> too much glare. The easiest way to keep glare from happening is to not have
> the sun directly "across" the target from your camera position.
>
I shot the targets in my studio. I measured very accurate triangles
at 45 degrees for the lights, target and camera.
The target I tried to keep perfectly vertical, I even used a spirit
level. It was off vertical by only a bit.
I shot the target with the image sensor on the camera parallel to the
target.
I will restudy your arrangement and once I undestand it fully I will
see if I can apply it.
Of note is that the version of ACR is newer and not compatible with
the script, you get the warning, but it runs anyway
On Jun 18, 5:08 am, "Steve Sprengel" <st...@sprengels.com> wrote:
> The calibration sliders the script produced for you are suspiciously large.
> To see if I get the same thing, I have started re-running the script on the
> H-software DNG you provided but my computer is about 6 times slower than
> yours so it will likely be the morning before I have something to compare to
> yours.
>
> I already tried running the script on the 3FR but Photoshop keeps getting an
> error a few iterations into the run. I am using 32-bit XP on a single-core
> machine with a bit more than 1 GB of memory available to Photoshop, but
> perhaps the size of this RAW is just too large for whatever the script is
> trying to do with it. I would guess a 64-bit version of PS wouldn't have
> the same sort of problem.
>
> Originally I applied the slider values you communicated to the .3FR file and
> Imatest showed significant oversaturation for most of the colors, but now I
> see that the slider values are from running on the H-software DNG, so I need
> to re-do the Imatest but the computer is busy doing the script for the next
> few hours so that will also be in tomorrow. It's possible that the .3FR
> and the .DNG will have identical errors, but until I try, I don't know,
> because I'm not sure what is different about the H-software version of the
> DNG from an Adobe software version.
The .3FR and .DNG files are theoretically different.
> You should also reshoot the CC24 RAW so that is isn't overexposed. The DNG
> Profile Editor reports that the three yellow squares (I think) and the white
> square have at least one channel clipped to 255, so I have been unable to
> use that tool to create a camera profile, either. If you reshoot the CC24
> dimmer, post another link to it, along with the H-version of the DNG. I
> almost always just shoot my target handheld in P mode, and don't have an
> exposure problem. If you have a gray-card you can set the exposure in
> M-mode using that. Green grass might be a substitute for a gray card to set
> the exposure with, although if it is well-fertilized, dark grass you might
> shoot at EV -0.5 to make sure it also doesn't overexpose.
>
and used the best take. I used Manual mode with a tripod.
Even that, I agree, some patches were slightly overexposed. I lowered
the exposure in ACR and the readings were all fine.
Is this not good enough?
> Imatest also complained that there is a vertical gradient (darker or lighter
> at the top that then bottom, not sure which), and maybe you can guess why
> that is, based on the way it is tilted and what is behind you. You might
> also try making it more of a 45-degree angle, pointed up a bit more at the
> sky, and shoot more down on it behind careful not to create a shadow of any
> kind on it. If I'm at a location is relatively clear of trees and
> buildings except along the horizon, I actually put it flat on the ground,
> sidewalk, driveway or street, then stand back a ways and zoom in quite a
> bit, and shoot the target at an angle from 45 to 60, with the wider edge
> about 3/4 of the way across the frame. Because you define the four corners
> separately it doesn't matter that it isn't parallel with the camera. You
> can think of what would be visible from the camera's point of view if the
> target was a mirror, and make sure that mostly only sky would be in that
> imagined reflection. Don't have the sun glaring off of it, too much,
> either. You can judge this by if the black plastic between the squares has
> too much glare. The easiest way to keep glare from happening is to not have
> the sun directly "across" the target from your camera position.
>
at 45 degrees for the lights, target and camera.
The target I tried to keep perfectly vertical, I even used a spirit
level. It was off vertical by only a bit.
I shot the target with the image sensor on the camera parallel to the
target.
I will restudy your arrangement and once I undestand it fully I will
see if I can apply it.
> ...
>
> read more »- Hide quoted text -
>
>
> read more »- Hide quoted text -
>
steve sprengel
Jun 18, 2009, 5:02:47 PM6/18/09
to acrcal...@googlegroups.com
The DNG Profile Editor does not care about "adjustments" you've made to the file via ACR. It is trying to calculate the underlying camera profile for the RAW data directly from the camera, not a set of calibration-slider adjustments to the existing profile and toning settings like the script is doing. If you bracketed the exposures, then just upload a darker version if you have one.
Of course you are using your studio setup and not sunlight outside...I realized this after I looked at your DNG, again, after the script was finished. Here are my revised suggestions for lighting in terms of a two-light setup that I'm imagining you're using:
Since Imatest complained about a vertical gradient and the white-patch is overexposed, I think the lights may be too close to the CC24 and moving things further away should even out the illumination more. My only experience is with non-studio lighting, so I am just guessing what is wrong. If they are already as far away as possible, then try to put them more on the same plane as the camera-to-target vector--more down to the level of the camera, instead of overhead or down low, depending on where they were for the first RAW.
--
Looking at the color-error plots I have had a chance to run, so far -- see attachments, it seems that:
--
I got baslcally the same large calibration-slider corrections when I ran the script on the H-version DNG as you did, so there's nothing wrong with how you specified the path or ran the script.
Can you explain how this DNG is created and if there are any options that would change how saturated it looks, or what color-profile it is using?
What is the purpose or the advantage of the H-version DNG over the .3FR file or a DNG created by the beta DNG Converter that is compatibel wtih ACR 5.4 RC?
Of course you are using your studio setup and not sunlight outside...I realized this after I looked at your DNG, again, after the script was finished. Here are my revised suggestions for lighting in terms of a two-light setup that I'm imagining you're using:
Since Imatest complained about a vertical gradient and the white-patch is overexposed, I think the lights may be too close to the CC24 and moving things further away should even out the illumination more. My only experience is with non-studio lighting, so I am just guessing what is wrong. If they are already as far away as possible, then try to put them more on the same plane as the camera-to-target vector--more down to the level of the camera, instead of overhead or down low, depending on where they were for the first RAW.
--
Looking at the color-error plots I have had a chance to run, so far -- see attachments, it seems that:
- The original .3FR (A000379) has the least color error.
- The H-version DNG (A_00739) is the worst and I am suspicious of its embedded profile or ACR's interpretation of it.
- The script-calibrated H-version DNG is close on some colors but the yellows are very oversaturated (blondes too yellow) and the purples undersaturated. This is due to the script applying large corrections but only looking at the Red, Green and Blue patches, and ignoring the Yellow and Cyan or Magenta colors so they tend to get worse instead of better.
--
I got baslcally the same large calibration-slider corrections when I ran the script on the H-version DNG as you did, so there's nothing wrong with how you specified the path or ran the script.
Can you explain how this DNG is created and if there are any options that would change how saturated it looks, or what color-profile it is using?
What is the purpose or the advantage of the H-version DNG over the .3FR file or a DNG created by the beta DNG Converter that is compatibel wtih ACR 5.4 RC?
Steve Sprengel
Jun 18, 2009, 5:33:22 PM6/18/09
to acrcal...@googlegroups.com
This arrived at my home e-mail with a blank headers, so
I am going to send it again, via my home e-mail, instead of Gmail.
Otherwise it is unchanged from the Gmail version so it may show up,
twice...
===
Antonio
Jun 19, 2009, 6:04:50 AM6/19/09
to AcrCalibrator
On Jun 18, 10:02 pm, steve sprengel <steve.spren...@gmail.com> wrote:
> The DNG Profile Editor does not care about "adjustments" you've made to the
> file via ACR. It is trying to calculate the underlying camera profile for
> the RAW data directly from the camera, not a set of calibration-slider
> adjustments to the existing profile and toning settings like the script is
> doing. If you bracketed the exposures, then just upload a darker version if
> you have one.
> Of course you are using your studio setup and not sunlight outside...I
> realized this after I looked at your DNG, again, after the script was
> finished. Here are my revised suggestions for lighting in terms of a
> two-light setup that I'm imagining you're using:
>
> Since Imatest complained about a vertical gradient and the white-patch is
> overexposed, I think the lights may be too close to the CC24 and moving
> things further away should even out the illumination more. My only
> experience is with non-studio lighting, so I am just guessing what is
> wrong. If they are already as far away as possible, then try to put them
> more on the same plane as the camera-to-target vector--more down to the
> level of the camera, instead of overhead or down low, depending on where
> they were for the first RAW.
as far a possible. The lights are about 3.8 meters apart. I can
arrange something bigger but will take time.
> Looking at the color-error plots I have had a chance to run, so far -- see
> attachments, it seems that:
>
> 2. The H-version DNG (A_00739) is the worst and I am suspicious of its
> embedded profile or ACR's interpretation of it.
> 3. The script-calibrated H-version DNG is close on some colors but the
> yellows are very oversaturated (blondes too yellow) and the purples
> undersaturated. This is due to the script applying large corrections but
> only looking at the Red, Green and Blue patches, and ignoring the Yellow and
> Cyan or Magenta colors so they tend to get worse instead of better.
>
I agree fully, the script calibration produces very saturated yellows.
> undersaturated. This is due to the script applying large corrections but
> only looking at the Red, Green and Blue patches, and ignoring the Yellow and
> Cyan or Magenta colors so they tend to get worse instead of better.
>
Your explanation makes sense and seems to conclude that the script
calibration is not very helpful, since missing out Yellow, Cyan and
Magenta is a major defficiency.
> I would like to try using the DNG Profile Editor on both the .3FR and the
> H-version DNG if you can upload a darker one. I am hoping that you can get
> a decent profile for the H-version DNG and can ignore the embedded profile,
> altogether.
will try to upload a darker one this weekend.
> I got baslcally the same large calibration-slider corrections when I ran the
> script on the H-version DNG as you did, so there's nothing wrong with how
> you specified the path or ran the script.
>
> Can you explain how this DNG is created and if there are any options that
> would change how saturated it looks, or what color-profile it is using?
It prides itself in being the best in image purity.
In addition to the usual RAW image adjustments (White balance,
exposure, sharpening...) it incorporates proprietary lens corrections
for Chromatic aberration, lens distortion, Vigneting and Moiré.
As far as I know the .DNG exported from Phocus from their .3FR RAW
captures should have the same color information, especially since
there are no lens corrections to make when shooting the target.
> What is the purpose or the advantage of the H-version DNG over the .3FR file
> or a DNG created by the beta DNG Converter that is compatibel wtih ACR 5.4
> RC?
>
processed by any other software (except as below)
.DNG files created by Hasselblad Phocus from their .3FR or .FFF files
can incorporate their lens adjustmens
The advantage of ACR 5.4 RC is simplicity of the workflow. One can
process the .3FR files directly on ACR 5.4 RC without going through
Phocus (.FFF cannot be processed by ACR 5.4 RC)
The downside is that , as far as I know, the lens corrections by
Hasselblad produce better results, but this is a small price to pay in
most cases.
> A0000739_nocal_colorerror.png
> 57KViewDownload
>
> A_00739_cal_colorerror.png
> 65KViewDownload
>
> A_00739_nocal_colorerror.png
> 66KViewDownload
Antonio
Jun 19, 2009, 9:11:35 AM6/19/09
to AcrCalibrator
Hello Steven,
Please find the darker .3FR here:
http://www.zshare.net/download/6158253397569904/
Please find the .DNG file created by Hasselblad Phocus from the above
file
http://www.zshare.net/download/61583167553b8d50/
The sidecar file with the white balance is
http://www.zshare.net/download/615832411e7b8be7/
The white balance from the WhiBal card that Photoshop shows is 5300 K
(exactly as the specs for the Bowens lights I am using), -7 tint
I did separate the lights as much as possible, remeasured and
readjusted the height of lights, camera and targets so that they are
now in a more even horizontal plane.
I took care the target is as vertical as possible.
I hope the shadows from the cardboard holder on the white and black
patches do not interfere. It is quite difficult to position the CC
card as needed with so small margins.
> > 66KViewDownload- Hide quoted text -
Please find the darker .3FR here:
http://www.zshare.net/download/6158253397569904/
Please find the .DNG file created by Hasselblad Phocus from the above
file
http://www.zshare.net/download/61583167553b8d50/
The sidecar file with the white balance is
http://www.zshare.net/download/615832411e7b8be7/
The white balance from the WhiBal card that Photoshop shows is 5300 K
(exactly as the specs for the Bowens lights I am using), -7 tint
I did separate the lights as much as possible, remeasured and
readjusted the height of lights, camera and targets so that they are
now in a more even horizontal plane.
I took care the target is as vertical as possible.
I hope the shadows from the cardboard holder on the white and black
patches do not interfere. It is quite difficult to position the CC
card as needed with so small margins.
Steve Sprengel
Jun 20, 2009, 8:33:59 PM6/20/09
to acrcal...@googlegroups.com
I meant to say that the lights should be further
from the target in that they be further towards or behind the camera, not
further to the sides, so that the diffuse circles from each light are
overlapping a bit more and evening out the illumination.
===
With this reshot target, DNG Profile Editor does not complain about the white target being clipped, and Imatest is not complaining about a horizontal or vertical gradient, but there are shadows from the cardboard on the sides of the white and black patches, that restrict the area I can sample with Imatest, but it's probably ok. In the future try to avoid any shadows as well as any non-neutral reflection from below the patches up onto them, like the cardboard is likely doing, perhaps making the neutrals slightly more brownish.
The DNG Profile Editor can create profiles from both the Phocus DNG and the 3FR once it is converted to a DNG via the Adobe DNG Converter.
Attached are the Imatest color-error results from both the 3FR and Phocus DNG and with respect to both the original (_Hprofile) and custom-profile (_CustProfile) created by the DNG Profile Editor.
In general, the overall color-error of the original profiles is about twice that of that of the DNG Profile Editor-produced profiles.
===
I could attach the DNG Profile Editor profiles I've created for you to use, but ACR doesn't "see" them, necessarily, unless you install them using the DNG Profile Editor, itself, using the File / Export ... profile function, so you should create your own profiles with the DNG Profile Editor, and let it install them on your system.
The steps I followed to do this are:
===
With this reshot target, DNG Profile Editor does not complain about the white target being clipped, and Imatest is not complaining about a horizontal or vertical gradient, but there are shadows from the cardboard on the sides of the white and black patches, that restrict the area I can sample with Imatest, but it's probably ok. In the future try to avoid any shadows as well as any non-neutral reflection from below the patches up onto them, like the cardboard is likely doing, perhaps making the neutrals slightly more brownish.
The DNG Profile Editor can create profiles from both the Phocus DNG and the 3FR once it is converted to a DNG via the Adobe DNG Converter.
Attached are the Imatest color-error results from both the 3FR and Phocus DNG and with respect to both the original (_Hprofile) and custom-profile (_CustProfile) created by the DNG Profile Editor.
In general, the overall color-error of the original profiles is about twice that of that of the DNG Profile Editor-produced profiles.
===
I could attach the DNG Profile Editor profiles I've created for you to use, but ACR doesn't "see" them, necessarily, unless you install them using the DNG Profile Editor, itself, using the File / Export ... profile function, so you should create your own profiles with the DNG Profile Editor, and let it install them on your system.
The steps I followed to do this are:
I first opened the DNG in ACR and set the WB using
the eyedropper on the white or next-to-white color-patch. I also went into
the Curve tab and choose Linear along with making sure the Camera Profile was
set to Hasselblad or Embedded. I don't know if any of this makes any
difference other than when I try to visualize the target in the DNG Profile
Editor, using the menu item Options / Apply ACR Settings.
Now here are my steps using the
DNGProfileEditor.exe that you've previously downloaded from http://labs.adobe.com/wiki/index.php/DNG_Profiles
File / Open the Phocus DNG of the CC24 target.
Click on the Chart tab at the right.
Drag the little circles to the center of their
respective corner patches.
Make sure it says Both Color Tables
Click on Create Profile.
Choose the Options tab
Type in a name for the profile like Phocus DNG
- Studio 5550K
Use File / Export... and save it using
the generated name, which is a combination of the camera-model and the
profile-name you typed in Options.
If you want to create a
profile for your 3FR, then you will need to convert it to a DNG using the Adobe
DNG Converter, first.
Because you've used "both color tables" with one DNG in
5500K lighting, this will only be valid for studio and sunlight usage.
===
One thing I noticed that
I'm not sure about, is that the one Hasselblad profile that comes with ACR says
Coated as part of its name, while the DNG Profile Editor says Uncoated when
opening the Phocus DNG. Is the term "coated" vs "uncoated" about different
camera sensor, or some indication of whether Phocus has manipulated the RAW
file.
On my Vista system, the Adobe-supplied camera
profiles are located in:
C:\ProgramData\Adobe\CameraRaw\CameraProfiles\Adobe
Standard
And any profiles that the DNG Profile Editor
installs are in:
C:\Users\<user
name>\AppData\Roaming\Adobe\CameraRaw\CameraProfiles
where this directory will also contain an
"index.dat" file that apparently is updated to include the names of the new
profiles you've created and exported. It seems to me that if you
just copy profiles to this directory without this index.dat existing, then they
don't appear to be seen by ACR and the only way to create the index.dat is to
export the profiles with the DNG Profile Editor. Perhaps this index.dat is
created by another process, and it is only coincidence but that is what seemed
to be happening to me, earlier, when I was creating profiles.
XP will have similar paths under C:\Documents and Settings.
--------------------------------------------------
From: "Antonio" <ans...@gmail.com>
Sent: Friday, June 19, 2009 8:11 AM
Antonio
Jun 22, 2009, 11:59:26 AM6/22/09
to AcrCalibrator
Hello Steve,
I am much indebted again, for your great help.
Please find my comments below.
On Jun 21, 1:33 am, "Steve Sprengel" <st...@sprengels.com> wrote:
> I meant to say that the lights should be further from the target in that they be further towards or behind the camera, not further to the sides, so that the diffuse circles from each light are overlapping a bit more and evening out the illumination.
I understand what you mean now. I placed lights and target at the
corners of a bigger 90 degree triangle. Lights are further apart from
the target. Although not exactly what you meant, I think the
ilumination was even enough.
> ===
> With this reshot target, DNG Profile Editor does not complain about the white target being clipped, and Imatest is not complaining about a horizontal or vertical gradient, but there are shadows from the cardboard on the sides of the white and black patches, that restrict the area I can sample with Imatest, but it's probably ok. In the future try to avoid any shadows as well as any non-neutral reflection from below the patches up onto them, like the cardboard is likely doing, perhaps making the neutrals slightly more brownish.
>
I agree, I should avoid the shadows.
> The DNG Profile Editor can create profiles from both the Phocus DNG and the 3FR once it is converted to a DNG via the Adobe DNG Converter.
>
> Attached are the Imatest color-error results from both the 3FR and Phocus DNG and with respect to both the original (_Hprofile) and custom-profile (_CustProfile) created by the DNG Profile Editor.
>
> In general, the overall color-error of the original profiles is about twice that of that of the DNG Profile Editor-produced profiles.
> ===
> I could attach the DNG Profile Editor profiles I've created for you to use, but ACR doesn't "see" them, necessarily, unless you install them using the DNG Profile Editor, itself, using the File / Export ... profile function, so you should create your own profiles with the DNG Profile Editor, and let it install them on your system.
>
I understand. That was exactly a question I had. I'll do that, and
thank you again, very much, for your detailed explanations.
> The steps I followed to do this are:
>
> I first opened the DNG in ACR and set the WB using the eyedropper on the white or next-to-white color-patch. I also went into the Curve tab and choose Linear along with making sure the Camera Profile was set to Hasselblad or Embedded. I don't know if any of this makes any difference other than when I try to visualize the target in the DNG Profile Editor, using the menu item Options / Apply ACR Settings.
>
> Now here are my steps using the DNGProfileEditor.exe that you've previously downloaded fromhttp://labs.adobe.com/wiki/index.php/DNG_Profiles
the Hasselblad literature.
> ...
>
> read more »
>
> A0000744_CustProfile_colorerror.png
> 65KViewDownload
>
> A0000744_Hprofile_colorerror.png
> 65KViewDownload
>
> A_00744_CustProfile_colorerror.png
> 65KViewDownload
>
> A_00744_Hprofile_colorerror.png
> 65KViewDownload- Hide quoted text -
I am much indebted again, for your great help.
Please find my comments below.
On Jun 21, 1:33 am, "Steve Sprengel" <st...@sprengels.com> wrote:
> I meant to say that the lights should be further from the target in that they be further towards or behind the camera, not further to the sides, so that the diffuse circles from each light are overlapping a bit more and evening out the illumination.
corners of a bigger 90 degree triangle. Lights are further apart from
the target. Although not exactly what you meant, I think the
ilumination was even enough.
> ===
> With this reshot target, DNG Profile Editor does not complain about the white target being clipped, and Imatest is not complaining about a horizontal or vertical gradient, but there are shadows from the cardboard on the sides of the white and black patches, that restrict the area I can sample with Imatest, but it's probably ok. In the future try to avoid any shadows as well as any non-neutral reflection from below the patches up onto them, like the cardboard is likely doing, perhaps making the neutrals slightly more brownish.
>
> The DNG Profile Editor can create profiles from both the Phocus DNG and the 3FR once it is converted to a DNG via the Adobe DNG Converter.
>
> Attached are the Imatest color-error results from both the 3FR and Phocus DNG and with respect to both the original (_Hprofile) and custom-profile (_CustProfile) created by the DNG Profile Editor.
>
> In general, the overall color-error of the original profiles is about twice that of that of the DNG Profile Editor-produced profiles.
> ===
> I could attach the DNG Profile Editor profiles I've created for you to use, but ACR doesn't "see" them, necessarily, unless you install them using the DNG Profile Editor, itself, using the File / Export ... profile function, so you should create your own profiles with the DNG Profile Editor, and let it install them on your system.
>
thank you again, very much, for your detailed explanations.
> The steps I followed to do this are:
>
> I first opened the DNG in ACR and set the WB using the eyedropper on the white or next-to-white color-patch. I also went into the Curve tab and choose Linear along with making sure the Camera Profile was set to Hasselblad or Embedded. I don't know if any of this makes any difference other than when I try to visualize the target in the DNG Profile Editor, using the menu item Options / Apply ACR Settings.
>
>
> File / Open the Phocus DNG of the CC24 target.
> Click on the Chart tab at the right.
> Drag the little circles to the center of their respective corner patches.
> Make sure it says Both Color Tables
> Click on Create Profile.
> Choose the Options tab
> Type in a name for the profile like Phocus DNG - Studio 5550K
> Use File / Export... and save it using the generated name, which is a combination of the camera-model and the profile-name you typed in Options.
>
> If you want to create a profile for your 3FR, then you will need to convert it to a DNG using the Adobe DNG Converter, first.
>
> Because you've used "both color tables" with one DNG in 5500K lighting, this will only be valid for studio and sunlight usage.
> ===
> One thing I noticed that I'm not sure about, is that the one Hasselblad profile that comes with ACR says Coated as part of its name, while the DNG Profile Editor says Uncoated when opening the Phocus DNG. Is the term "coated" vs "uncoated" about different camera sensor, or some indication of whether Phocus has manipulated the RAW file.
>
I am not aware of the term "coated" as having a specific meaning in
> File / Open the Phocus DNG of the CC24 target.
> Click on the Chart tab at the right.
> Drag the little circles to the center of their respective corner patches.
> Make sure it says Both Color Tables
> Click on Create Profile.
> Choose the Options tab
> Type in a name for the profile like Phocus DNG - Studio 5550K
> Use File / Export... and save it using the generated name, which is a combination of the camera-model and the profile-name you typed in Options.
>
> If you want to create a profile for your 3FR, then you will need to convert it to a DNG using the Adobe DNG Converter, first.
>
> Because you've used "both color tables" with one DNG in 5500K lighting, this will only be valid for studio and sunlight usage.
> ===
> One thing I noticed that I'm not sure about, is that the one Hasselblad profile that comes with ACR says Coated as part of its name, while the DNG Profile Editor says Uncoated when opening the Phocus DNG. Is the term "coated" vs "uncoated" about different camera sensor, or some indication of whether Phocus has manipulated the RAW file.
>
the Hasselblad literature.
>
> read more »
>
> A0000744_CustProfile_colorerror.png
> 65KViewDownload
>
> A0000744_Hprofile_colorerror.png
> 65KViewDownload
>
> A_00744_CustProfile_colorerror.png
> 65KViewDownload
>
> A_00744_Hprofile_colorerror.png
> 65KViewDownload- Hide quoted text -
Steve Sprengel
Jun 22, 2009, 12:20:42 PM6/22/09
to acrcal...@googlegroups.com
If you are unaware of any other meaning, I expect
"Coated" and "Uncoated" refers to the Carl-Zeiss T* coating on newer
lenses:
http://www.pebbleplace.com/Personal/Contax_db_Coatings.html
http://www.pebbleplace.com/Personal/Contax_db_Coatings.html
and since the coating influences colors, you'd want
a profile that took this into account, which is another reason to create your
own custom profile perhaps for specific lenses you work with.
--------------------------------------------------
From: "Antonio" <ans...@gmail.com>
Sent: Monday, June 22, 2009 10:59 AM
Juan
Jul 8, 2009, 1:11:38 PM7/8/09
to AcrCalibrator
HI,
I have CCSG as well as CC24 and I recently shot CCSG using 2 softboxes
at 45 degrees - as recommended.
th white patches were well inside the allowed values (I believe 235)
and the black were under 20 (actually around 7). The max deviation in
the white patches was 2 or 3, just as for the middle grays and blacks.
So I considered I had the BEST shot of CCSG I had made in my life.
with these high expectations, I created a profile using Profile Maker
5.0.5 and the raw converter CaptureOne which is able to create a 16
bit TIFF with no color correction (meaning an absence of profile) .
The results are discouraging. The raw images are open in CaptureOne
and the new profile is assigned to them - they become too bright and
with low contrast. I don't know if the no-profile profile called No
color conversion is truly neutral, if it is showing the exact image as
seen by the camera. Do you know of any software that would display a
RAW file without a profile applied? It seems to me that the error must
lie here, because what colors can be associated with a RAW file by
itself, since a profile is what links colors to numbers in the RAW
file, isn't it?
Thanks,
Juan
I have CCSG as well as CC24 and I recently shot CCSG using 2 softboxes
at 45 degrees - as recommended.
th white patches were well inside the allowed values (I believe 235)
and the black were under 20 (actually around 7). The max deviation in
the white patches was 2 or 3, just as for the middle grays and blacks.
So I considered I had the BEST shot of CCSG I had made in my life.
with these high expectations, I created a profile using Profile Maker
5.0.5 and the raw converter CaptureOne which is able to create a 16
bit TIFF with no color correction (meaning an absence of profile) .
The results are discouraging. The raw images are open in CaptureOne
and the new profile is assigned to them - they become too bright and
with low contrast. I don't know if the no-profile profile called No
color conversion is truly neutral, if it is showing the exact image as
seen by the camera. Do you know of any software that would display a
RAW file without a profile applied? It seems to me that the error must
lie here, because what colors can be associated with a RAW file by
itself, since a profile is what links colors to numbers in the RAW
file, isn't it?
Thanks,
Juan
steve sprengel
Jul 8, 2009, 3:34:20 PM7/8/09
to acrcal...@googlegroups.com
Am I right in thinking you are trying to create a camera profile for use with C1 using ProfileMaker--something analogus to using DNG Profile Editor with Adobe Camera RAW but it's not working out? Is the process you're following documented to work and is the recommended way to make a profile or are you just attempting a process that you're just assuming should work? I ask this because I don't use ProfileMaker nor have used C1 other than a trial period some years ago so I don't have specific experience with the procedure you're trying nor am I sure whether what you are attempting to do is anything meaningful or not although without knowing any better I'd think it should. My direct experience is limited to ACR/LR and the scripts and DNG Profile Editor.
One question I have, right off, is what is the input to ProfileMaker when you create your profile? Is it the RAW file, or the TIFF from C1 w/o a profile? I would expect the latter but you mention the two products in the opposite order so I'm not sure.
I do know that the data in a RAW file when viewed by itself looks very dim. This is because the response of your eye is not nearly linear, so typically a gamma curve has to be applied to the RAW data to make it look similar to how our eyes see things. I don't know if a color-profile is enough to apply this gamma curve, or if the gamma-curve would be done before or after making a profile in ProfileMaker and how one would even apply a gamma curve to the data.
You talk about C1 being able to create a 16-bit TIFF w/o a color profile, and then ask if it is possible to see the RAW file w/o a color profile...isn't the uncalibrated 16-bit TIFF output of C1 what you are asking for?
Or, if you mean RAWer than the C1 TIFF, then understand that a RAW file is only one-color-per-pixel and is monochrome data of whatever light itensity was behind each of the color-filters in the camera, so it is not clear what you expect to see when looking at the original RAW data without a color profile applied. It is possible to visualize a RAW file before it has been converted to a 3-colors-per-pixel RGB bitmap (a TIFF or JPG), and you do this by starting with the RAW data converted to a monochrome TIFF and then using masks in photoshop to assign green to every other pixel and red to every fourth pixel and blue to the other fourth of the pixels--assigning colors to each pixel based on what color the filter was in front of that photosite in the camera. This results in something that sortof looks right until you zoom in and see that things are made up of red, green and blue dots with there being twice as many green dots. I did this probably 10 yeras ago, once, for a D70 image that had bad mosaic errors, but don't recall the software I used to do it...perhaps ImagesPlus, a piece of astrophotography software that works with the monochorme RAW data to remove hot pixels and reduce noise before the images are converted to color images.
To create an RGB TIFF that looks reasonable from RAW data, you have to do two things:
1) Take the matrix of one-color-per-pixel data interpolate the remaining 2 colors at each pixel location from neighboring pixels using information about the transmissivity of the various color filters in the particular camera sensor--a process called demosaicing, and
2) Apply a gamma curve to the data which brightens the darker pixels more than the brighter pixels to simulate how the eye sees things.
As to which of these happens first, it's hard to say, and maybe they happen simultaneously, but it is the job of the RAW converter to figure out the best RGB value for each pixel based on the original data only having R or G or B in any one location.
At least the transmissivity of the red, green, and blue filters is something that makes each sensor unique and is why you have to have a camera profile. Also if the sensor, itself, is not linear, then the shape of the gamma curve would be slightly different for each sensor and presumably the camera-profile takes that into account.
It is not clear from the process you describe if the uncalibrated 16-bit TIFF from C1 would have a gamma-curve applied or not. I could imagine the results you are getting could be that one piece of software is assuming the gamma curve has been applied and one piece of software is assuming it has not and needs to have it applied, and so the profile is being created either with two-gamma curves applied or non, instead of just one time.
--
To get an idea of what RAW data looks like you can look at what is captioned "RAW image" on this page about the demosacing algorithm in a digital camera sensor:
http://electronics.howstuffworks.com/digital-camera5.htm
Of course this RAW image already has pseudo-color applied to each location of either red, green or blue, but you can see how different it looks from the resulting demosaiced image you see in the animation after you click the Begin Demosaicing Algorithm button.
==
Although it's pretty technical and has a particular conclusion it is trying to support, here is a page that discusses some of the issues involved in translating digital-camera sensor values into something that looks "right" compared to what our eyes see:
http://21stcenturyshoebox.com/essays/color_reproduction.html
If you look at the Image Processing Pipeline figure, you'll see that the Calibration step is very near the beginning while the Gamma Encoding step is near the end, so that is why I wonder if there is a gamma-curve problem with the image you don't think looks right.
As to how applicable the ColorCheckerSG is creating profiles, read the last section at the very bottom of this page:
http://www.babelcolor.com/main_level/ColorChecker.htm
One question I have, right off, is what is the input to ProfileMaker when you create your profile? Is it the RAW file, or the TIFF from C1 w/o a profile? I would expect the latter but you mention the two products in the opposite order so I'm not sure.
I do know that the data in a RAW file when viewed by itself looks very dim. This is because the response of your eye is not nearly linear, so typically a gamma curve has to be applied to the RAW data to make it look similar to how our eyes see things. I don't know if a color-profile is enough to apply this gamma curve, or if the gamma-curve would be done before or after making a profile in ProfileMaker and how one would even apply a gamma curve to the data.
You talk about C1 being able to create a 16-bit TIFF w/o a color profile, and then ask if it is possible to see the RAW file w/o a color profile...isn't the uncalibrated 16-bit TIFF output of C1 what you are asking for?
Or, if you mean RAWer than the C1 TIFF, then understand that a RAW file is only one-color-per-pixel and is monochrome data of whatever light itensity was behind each of the color-filters in the camera, so it is not clear what you expect to see when looking at the original RAW data without a color profile applied. It is possible to visualize a RAW file before it has been converted to a 3-colors-per-pixel RGB bitmap (a TIFF or JPG), and you do this by starting with the RAW data converted to a monochrome TIFF and then using masks in photoshop to assign green to every other pixel and red to every fourth pixel and blue to the other fourth of the pixels--assigning colors to each pixel based on what color the filter was in front of that photosite in the camera. This results in something that sortof looks right until you zoom in and see that things are made up of red, green and blue dots with there being twice as many green dots. I did this probably 10 yeras ago, once, for a D70 image that had bad mosaic errors, but don't recall the software I used to do it...perhaps ImagesPlus, a piece of astrophotography software that works with the monochorme RAW data to remove hot pixels and reduce noise before the images are converted to color images.
To create an RGB TIFF that looks reasonable from RAW data, you have to do two things:
1) Take the matrix of one-color-per-pixel data interpolate the remaining 2 colors at each pixel location from neighboring pixels using information about the transmissivity of the various color filters in the particular camera sensor--a process called demosaicing, and
2) Apply a gamma curve to the data which brightens the darker pixels more than the brighter pixels to simulate how the eye sees things.
As to which of these happens first, it's hard to say, and maybe they happen simultaneously, but it is the job of the RAW converter to figure out the best RGB value for each pixel based on the original data only having R or G or B in any one location.
At least the transmissivity of the red, green, and blue filters is something that makes each sensor unique and is why you have to have a camera profile. Also if the sensor, itself, is not linear, then the shape of the gamma curve would be slightly different for each sensor and presumably the camera-profile takes that into account.
It is not clear from the process you describe if the uncalibrated 16-bit TIFF from C1 would have a gamma-curve applied or not. I could imagine the results you are getting could be that one piece of software is assuming the gamma curve has been applied and one piece of software is assuming it has not and needs to have it applied, and so the profile is being created either with two-gamma curves applied or non, instead of just one time.
--
To get an idea of what RAW data looks like you can look at what is captioned "RAW image" on this page about the demosacing algorithm in a digital camera sensor:
http://electronics.howstuffworks.com/digital-camera5.htm
Of course this RAW image already has pseudo-color applied to each location of either red, green or blue, but you can see how different it looks from the resulting demosaiced image you see in the animation after you click the Begin Demosaicing Algorithm button.
==
Although it's pretty technical and has a particular conclusion it is trying to support, here is a page that discusses some of the issues involved in translating digital-camera sensor values into something that looks "right" compared to what our eyes see:
http://21stcenturyshoebox.com/essays/color_reproduction.html
If you look at the Image Processing Pipeline figure, you'll see that the Calibration step is very near the beginning while the Gamma Encoding step is near the end, so that is why I wonder if there is a gamma-curve problem with the image you don't think looks right.
As to how applicable the ColorCheckerSG is creating profiles, read the last section at the very bottom of this page:
http://www.babelcolor.com/main_level/ColorChecker.htm
Reply all
Reply to author
Forward
0 new messages