Ad
Ad
Ad
Pages: « 1 [2] 3 »   Bottom of Page
Print
Author Topic: Camera Calibration for Copy Work  (Read 6144 times)
louoates
Sr. Member
****
Offline Offline

Posts: 768



WWW
« Reply #20 on: July 22, 2013, 05:19:26 PM »
ReplyReply

I appreciate you trying to get 100 perfect color every time and applaud your perseverance. Your particular market may demand exceedingly accurate reproductions. Thank goodness my copy work flow for artists were never that stringent. I promised 95% color accuracy for most colors and only best-possible approximations for florescent and rare-earth type paints. Maybe it was because I wasn't charging as much as others in the area. In the four years I was printing for 45-some artists (mostly for art show sales) I never had to re-shoot a painting because of artist dissatisfaction. In my case the strategy of under-promise and over-deliver worked well.
   I was also surprised at how many artists had trouble evaluating color values in their own work.
Logged
Tim Lookingbill
Sr. Member
****
Offline Offline

Posts: 1158



WWW
« Reply #21 on: July 22, 2013, 05:27:11 PM »
ReplyReply

Quote
I am not suggesting you need to do this testing for me, just that it doesn't really prove a point.

It proves that I get a better and quicker match with my setup with very little effort over yours which you seem to be having trouble with by the fact you're having to split hairs over two CCchart color patches that are off by the numbers instead of just getting the work done.

The real problem that you don't address is how you expect to get a visual match under the warm glow of halogen lighting while expecting the Lab numbers to be perfect which are derived and measured from a D50 light source.

Of course it would make sense you shouldn't get an exact CCchart match by the numbers which I have to admit I certainly don't get using the ACR 4.4 profile (in fact the numbers are far more off than yours using that profile), however, I do get a visual match with that particular photo which is the purpose of this whole process. For all intents and purposes I shouldn't get a visual match but I do.

What's going to happen when you go through all this trouble to get all the numbers perfect, but the appearance of the final copy doesn't match what you see due to the fact your eyes have adapted to that warm halogen glow which isn't a D50 light source? And filters add another level of complexity because you, the camera and PP software don't know the exact spectra formulation and its effect on your "variety" of halogen which there can be quite a few hues to contend with.

Not all what's called D50 looks the same between manufacturers of various color matching processes.
Logged
Iliah
Sr. Member
****
Offline Offline

Posts: 410


« Reply #22 on: July 22, 2013, 06:02:07 PM »
ReplyReply

> is there a second choice?
PhaseOne CaptureOne (60-day trial), RawTherapy (free), and pretty much any other raw converter from a third party. You can with some effort make Nikon View (free) or Nikon Capture (30-day trial) work with a custom profile as well. ColorChecker SG is a better target for profiling compared to DataColor's - more consistent and wider gamut. If you rotate the pola filter the profile can be off.
Logged
Iliah
Sr. Member
****
Offline Offline

Posts: 410


« Reply #23 on: July 22, 2013, 06:04:41 PM »
ReplyReply

Some brief how-to http://sail2ithaki.livejournal.com/187421.html
Logged
Iliah
Sr. Member
****
Offline Offline

Posts: 410


« Reply #24 on: July 22, 2013, 06:07:49 PM »
ReplyReply

> https://docs.google.com/file/d/0B-YoFPgh003pUzZwV1RIRWlPNVk/edit?usp=sharing

The shot is close to 1 stop underexposed.
Logged
RogerEle
Newbie
*
Offline Offline

Posts: 31


« Reply #25 on: July 22, 2013, 10:20:55 PM »
ReplyReply


Iliah, thanks for the input and the links

Regarding the linked image being underexposed - I just downloaded it (I am at home Smiley ), the lightest patch reads rgb 253, 253, 252 in Adobe(1998) - there is room for more exposure ....  Shocked ?  Darker tones are darker because of polarizing - would you expose for the midtone and the bring the light tones down with adjustments?

Thanks,
Roger
Logged
RogerEle
Newbie
*
Offline Offline

Posts: 31


« Reply #26 on: July 22, 2013, 10:28:54 PM »
ReplyReply

I appreciate you trying to get 100 perfect color every time and applaud your perseverance. Your particular market may demand exceedingly accurate reproductions. Thank goodness my copy work flow for artists were never that stringent. I promised 95% color accuracy for most colors and only best-possible approximations for florescent and rare-earth type paints. Maybe it was because I wasn't charging as much as others in the area. In the four years I was printing for 45-some artists (mostly for art show sales) I never had to re-shoot a painting because of artist dissatisfaction. In my case the strategy of under-promise and over-deliver worked well.
   I was also surprised at how many artists had trouble evaluating color values in their own work.

Hi louoates,

I too under promise and over deliver, as a conscious way of doing business!  We get happier clients and more referrals! Smiley I have a personality that always has to be tweaking something ... this is the focus of the moment ... hoping to bump what I do a little closer to perfection Smiley

Thanks
Logged
RogerEle
Newbie
*
Offline Offline

Posts: 31


« Reply #27 on: July 22, 2013, 10:49:18 PM »
ReplyReply

It proves that I get a better and quicker match with my setup with very little effort over yours which you seem to be having trouble with by the fact you're having to split hairs over two CCchart color patches that are off by the numbers instead of just getting the work done.

You are right, but from time to time I just need to prove to myself that I have it nailed as well as I can.  I live by the motto's ...
- It is what it is
- I do the best I can
- I don't know what I don't know

Quote
The real problem that you don't address is how you expect to get a visual match under the warm glow of halogen lighting while expecting the Lab numbers to be perfect which are derived and measured from a D50 light source.

Because I didn't know that was a problem (third motto), but now that I do I will test it with flash Smiley

Quote
Of course it would make sense you shouldn't get an exact CCchart match by the numbers which I have to admit I certainly don't get using the ACR 4.4 profile (in fact the numbers are far more off than yours using that profile), however, I do get a visual match with that particular photo which is the purpose of this whole process. For all intents and purposes I shouldn't get a visual match but I do.

I get a visual match on many of the images that I am copying - just seeing if I can up my percentages - every time that I have poked at this stuff in the past I have learned something that reinforces the third motto

Quote
What's going to happen when you go through all this trouble to get all the numbers perfect, but the appearance of the final copy doesn't match what you see due to the fact your eyes have adapted to that warm halogen glow which isn't a D50 light source? And filters add another level of complexity because you, the camera and PP software don't know the exact spectra formulation and its effect on your "variety" of halogen which there can be quite a few hues to contend with.

Not all what's called D50 looks the same between manufacturers of various color matching processes.

I don't understand the 'eyes have adapted to that warm halogen glow' part - I am not looking at the copy stand and lights while working on the computer - but regardless, if it is all messed up - that teaches something too - I will test the calibration with the flash instead of the halogen / 80a mix so that I 'for sure' know the difference!

Thanks!
Roger
Logged
Vladimirovich
Sr. Member
****
Offline Offline

Posts: 1320


« Reply #28 on: July 22, 2013, 11:11:25 PM »
ReplyReply

Regarding the linked image being underexposed - I just downloaded it (I am at home Smiley ), the lightest patch reads rgb 253, 253, 252 in Adobe(1998) - there is room for more exposure ....  Shocked ?
I think Iliah is talking about the raw data (check the patch with rawdigger) for the purpose of icc profile creation... you posted a target shot.

PS: for example Eric Chan for Adobe PE (and .dcp camera profiles) suggested to bracket exposure and use the most exposed raw (converted to DNG) that Adobe PE still accepts.
« Last Edit: July 22, 2013, 11:14:40 PM by Vladimirovich » Logged
MarkM
Sr. Member
****
Offline Offline

Posts: 327



WWW
« Reply #29 on: July 23, 2013, 03:32:54 AM »
ReplyReply

Roger,

I think you should hang on to that $400 until you've spent more time with what you have. If you are willing to get in and tweak the profiles, you can get a colorchecker arbitrarily close to published values with nothing more than free software. Whether this will actually be useful in practice with real-world images and workflow is debatable. But if you're like me you'll find it an interesting and educational process.

I spend a few minutes with you RAW file in the Adobe DNG profile Editor and was able to get all the patches within less than Delta-E 2 of published RGB numbers for the color checker. The profile alone can do this without touching a single slider in ACR (other than making sure your white balance is right, which, based on the 4th grey patch of your raw, I set to Temp: 4250, Tint: +26).

The most important part of the profile if you are going for numerical accuracy in my experience is getting the tone curve right. The default is almost always too contrasty and a linear leaves the shadows too dark. Once you get the tone curve right, many of the other colors fall in line. Depending on the character of your camera and light, you may need to push a few of the patches around a bit.

Personally, I don't think there's anything wrong with using hot lights and a cooling filter. The spectrum from that combination is going to be pretty decent especially when you have tools like ACR and DNG profiles at you disposal.

I've attached the processed raw, a diagonal cut comparison with published numbers—not perfect, but pretty good for a minimal effort, and the DNG profile I made/used.

Good luck.
« Last Edit: July 23, 2013, 03:54:14 AM by MarkM » Logged

opgr
Sr. Member
****
Offline Offline

Posts: 1125


WWW
« Reply #30 on: July 23, 2013, 06:01:01 AM »
ReplyReply

To me colormanagement and profiling is about consistency of results. Exact reproduction of a colorchecker is a side effect of that, it is not a goal in itself.

What currently happens is that a non-linear response is introduced at the start of a colormanagement chain which is entirely based on a linear response assumption. Using the DNG profiler to adjust the profile with all kinds of contorted hue shifts then is equivalent to "introducing errors" for the sake of reproducing a colorchecker correctly. It doesn't guarantee consistency of color at all, and, in fact, may simply enlarge the actual error of colorreproduction.

What needs to be done in my opinion is create the profile with the polarizer in "off mode" and specifically NOT try to adjust for individual colorchecker colors. i.e. get as close to a clean matrix profile as possible. With this profile, try to take a polarized shot and only adjust the contrast curve in LR and see if that gets the colors in place.

The unfortunate issue is that the colormanagement chain in LR is simply not transparent enough to know how to implement the "technically" correct steps, so that you can actually conclude that it does or does not work for your purposes. The technically correct steps for your workflow imo like are:

1. take a polarized shot of artwork
(introduces non-linearities in the form of contrast increase)

2. compensate for the contrast increase

3. apply a clean linear matrix profile
(based on a non-polarized shot of the colorchecker).

Logged

Regards,
Oscar Rysdyk
theimagingfactory
Vladimirovich
Sr. Member
****
Offline Offline

Posts: 1320


« Reply #31 on: July 23, 2013, 08:20:16 AM »
ReplyReply

What needs to be done in my opinion is create the profile with the polarizer in "off mode" and specifically NOT try to adjust for individual colorchecker colors. i.e. get as close to a clean matrix profile as possible.

you can strip LUTs from .dcp using SandyMc' dcptool... however if you are using Adobe PE - you are always using some base profile (so unless you create some genuinely new one /matrix wise/ with for example OEM Xrite software or if QPCard will release a plugin for XRite passport or by changing matrices manually) you are using Adobe's supplied matrices and then if you do not need LUT you do not need to shoot colochecker at all at that stage - just strip the details from Adobe's profile (Standard or one of OEM emulations) and get a clean base profile that is matrix only (with proper curve and BaselineExposureOffset as well)
Logged
Iliah
Sr. Member
****
Offline Offline

Posts: 410


« Reply #32 on: July 23, 2013, 08:21:35 AM »
ReplyReply

> Regarding the linked image being underexposed

It is a shot in 12 bit mode with 2660 max in it in green channel; while technically it should be around 4000 (2^12-1). You can open the shot with RawDigger and see for yourself. You can shoot a scene with a specular highlight in it (like from some metal ball) to determine the maximum for your camera.

Second, you filled the frame with the target, so you are making lighting task more difficult, and also get into any lens and sensor irregularities. Better to have the target in 1/3 of the frame.

Target is shot in sharp focus, it does not help too.

Target is not clean, readings will be off.

To equalize the light put some uniform white or grey sheet in the first shot, then use flat field feature of RawDigger to export CGATS corrected against light non-uniformity.

I would be glad to help you further if needed.
Logged
RogerEle
Newbie
*
Offline Offline

Posts: 31


« Reply #33 on: July 23, 2013, 12:22:24 PM »
ReplyReply

I spend a few minutes with you RAW file in the Adobe DNG profile Editor and was able to get all the patches within less than Delta-E 2 of published RGB numbers for the color checker. The profile alone can do this without touching a single slider in ACR (other than making sure your white balance is right, which, based on the 4th grey patch of your raw, I set to Temp: 4250, Tint: +26).

Thanks!

Quote
The most important part of the profile if you are going for numerical accuracy in my experience is getting the tone curve right. The default is almost always too contrasty and a linear leaves the shadows too dark. Once you get the tone curve right, many of the other colors fall in line.

Not so much numerical accuracy, but no surprises when copying - the numbers seemed a good place to to start - I won't be sweating the tiny differences that don't show up in actual use.
Thanks you for describing this so clearly! - it lets me know I wasn't off in what I was finding

Quote
Personally, I don't think there's anything wrong with using hot lights and a cooling filter. The spectrum from that combination is going to be pretty decent especially when you have tools like ACR and DNG profiles at you disposal.

Good to hear Smiley

Thanks! Mark,
Logged
RogerEle
Newbie
*
Offline Offline

Posts: 31


« Reply #34 on: July 23, 2013, 12:28:04 PM »
ReplyReply

To me colormanagement and profiling is about consistency of results. Exact reproduction of a colorchecker is a side effect of that, it is not a goal in itself.

What currently happens is that a non-linear response is introduced at the start of a colormanagement chain which is entirely based on a linear response assumption. Using the DNG profiler to adjust the profile with all kinds of contorted hue shifts then is equivalent to "introducing errors" for the sake of reproducing a colorchecker correctly. It doesn't guarantee consistency of color at all, and, in fact, may simply enlarge the actual error of colorreproduction.

What needs to be done in my opinion is create the profile with the polarizer in "off mode" and specifically NOT try to adjust for individual colorchecker colors. i.e. get as close to a clean matrix profile as possible. With this profile, try to take a polarized shot and only adjust the contrast curve in LR and see if that gets the colors in place.

The unfortunate issue is that the colormanagement chain in LR is simply not transparent enough to know how to implement the "technically" correct steps, so that you can actually conclude that it does or does not work for your purposes. The technically correct steps for your workflow imo like are:

1. take a polarized shot of artwork
(introduces non-linearities in the form of contrast increase)

2. compensate for the contrast increase

3. apply a clean linear matrix profile
(based on a non-polarized shot of the colorchecker).



Thanks Oscar!

I did photograph the CC target with the polarizer turned to the off position - I stopped test with it in off when I discovered the colors were close and I realized that I would never be actually using it in the off position - I will definitely give what you suggest a try!

Roger
Logged
RogerEle
Newbie
*
Offline Offline

Posts: 31


« Reply #35 on: July 23, 2013, 12:31:24 PM »
ReplyReply

you can strip LUTs from .dcp using SandyMc' dcptool... however if you are using Adobe PE - you are always using some base profile (so unless you create some genuinely new one /matrix wise/ with for example OEM Xrite software or if QPCard will release a plugin for XRite passport or by changing matrices manually) you are using Adobe's supplied matrices and then if you do not need LUT you do not need to shoot colochecker at all at that stage - just strip the details from Adobe's profile (Standard or one of OEM emulations) and get a clean base profile that is matrix only (with proper curve and BaselineExposureOffset as well)

Wow, I will have to come back to this after fooling with the recommended raw software - gives me a peek though, thanks!
Logged
RogerEle
Newbie
*
Offline Offline

Posts: 31


« Reply #36 on: July 23, 2013, 12:51:46 PM »
ReplyReply

> Regarding the linked image being underexposed

It is a shot in 12 bit mode with 2660 max in it in green channel; while technically it should be around 4000 (2^12-1). You can open the shot with RawDigger and see for yourself. You can shoot a scene with a specular highlight in it (like from some metal ball) to determine the maximum for your camera.

Second, you filled the frame with the target, so you are making lighting task more difficult, and also get into any lens and sensor irregularities. Better to have the target in 1/3 of the frame.

Target is shot in sharp focus, it does not help too.

Target is not clean, readings will be off.

To equalize the light put some uniform white or grey sheet in the first shot, then use flat field feature of RawDigger to export CGATS corrected against light non-uniformity.

I would be glad to help you further if needed.

Thanks Iliah! - I will be on vacation for a week and will load the software's to my laptop while relaxing - will see if I can photo new targets before I leave ...
Logged
RogerEle
Newbie
*
Offline Offline

Posts: 31


« Reply #37 on: July 23, 2013, 01:03:53 PM »
ReplyReply

Notes ... Smiley

I did shoot new targets yesterday before the latest replies.  Not knowing any better I have been exposing to place the exposure in the middle of the histogram. 

Here is the flash .nef

https://docs.google.com/a/eleakis.com/file/d/0B-YoFPgh003pS1hHQkhTc1BpZkU/edit?usp=sharing

It is definitely different from either of the tungsten/80a images (with polarizer on and off), will have to see how the profiles pan out but adds initial credence to the idea that tungsten with 80a is not equivalent to a clean D50 light source - will have to see Smiley

Here is the .nef with polarizer turned off, way under exposed by my current understanding

https://docs.google.com/a/eleakis.com/file/d/0B-YoFPgh003pdHZFUEpIbTN5Yjg/edit?usp=sharing

Thanks!
Roger
Logged
MarkM
Sr. Member
****
Offline Offline

Posts: 327



WWW
« Reply #38 on: July 23, 2013, 01:11:42 PM »
ReplyReply

Roger, it looks like those files are not publicly accessible.
Logged

Iliah
Sr. Member
****
Offline Offline

Posts: 410


« Reply #39 on: July 23, 2013, 01:16:42 PM »
ReplyReply

The idea is to profile for light. My setup includes Rosco #1991 or #1995, CTB, pola filter, and a 50% magenta filter on the halogen lights. This way I have full spectrum and equal exposure for all raw channels.

I seriously suggest looking at ColorChecker SG target for profiling, setting the exposure from SpyderCUBE or similar self-made, and having a flat grey surface for both profiling and repro work.
Logged
Pages: « 1 [2] 3 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad