Gamma converting monitor ICC profile to printer's profile

Moderator: jsachs

Post Reply
tomczak
Posts: 1367
Joined: April 25th, 2009, 12:56 am
What is the make/model of your primary camera?: Fuji X-E2
Contact:

Gamma converting monitor ICC profile to printer's profile

Post by tomczak »

Here is what I think, but I'm not sure.

In a standard monitor/working colour profile (e.g. sRGB) there are xTRC tags that define display gamma for this profile. They don't have LUTs.

Printing profile don't have xTRC tags, but typically have several LUTs for different rendering intents. The LUTs look more or less 'linear'

When converting the image data from sRGB colour space to Printer's colour space, are pixel values converted according to gamma in sRGB profile first, then those 'gamma-converted' values are subjected to printer's profile LUTs to calculate the RGB values in the printer profile (thus no need for the second gamma)? What happens if the target profile has no LUTs?

Here is the reason I'm asking: what defines the printer's gamma when using colour-managed workflow - it looks to me that it's the gamma defined in the working profile? Does it make any sense?

Cheers.
Maciej Tomczak
Phototramp.com
jsachs
Posts: 4203
Joined: January 22nd, 2009, 11:03 pm

Re: Gamma converting monitor ICC profile to printer's profile

Post by jsachs »

First the TRC flags in a monitor profile may either be a simple gamma value or a LUT (TRC stands for tone reproduction curve). sRGB actually uses a table because it treats the first few values different from a standard gamma curve although you would be hard-pressed to notice the difference.

Printer profiles are generally based on 3-D lookup tables instead of the 3 1-D lookup tables that monitor profiles usually have. There is much more overhead using a 3-D table since it required 3-D interpolation for each pixel which is an expensive operation, while 1-D table lookup is very fast. Printer profiles also support a set of 3 1-D lookup tables applied to the data before going through the 3-D table and another set applied to the data after the 3-D table. The 3 1-D tables applied before are usually used to adjust gamma and make sure grays reproduce as neutral.

Conversion from the working color space to the printer space is accomplished via an intermediate color space called the Profile Connection Space. At each step the gammas of the respective color spaces are taken into account.
Jonathan Sachs
Digital Light & Color
tomczak
Posts: 1367
Joined: April 25th, 2009, 12:56 am
What is the make/model of your primary camera?: Fuji X-E2
Contact:

Re: Gamma converting monitor ICC profile to printer's profile

Post by tomczak »

So what's the easiest way to find out what would a gamma (or similar piece-wise transfer curve) is used in a printer profile?

The reason for the question is this: I have a hard time to print shaddows properly on Durst Lambda. Even if the test prints are seen in really bright light, they seem to have blacks compressed and highlights extended unnecessarily. The midtone of the black to white ramp strip is much closer to black than it shoud be.

The test prints seems to indicate to me that the effective 'gamma' of the printer is much too high, comparing to the standard 2.2 gamma (or an sRGB piece-wise function really close to it) that the image pixel data are subjected to before being displayed on the monitor (and than the eyes reverses the process, stretching shadows and compressing highlights in their own complex way).

I don't do anything special here: images are processed in the working colour space with gamma 2.2 and then converted as 48bits files to the given printer/paper callibration ICC profile. But I don't know how to find out what 'gamma' does this profile applies to the image data (forget the gammut conversion) when the ink opacity is decided on while printing.

Cheers.
Maciej Tomczak
Phototramp.com
jsachs
Posts: 4203
Joined: January 22nd, 2009, 11:03 pm

Re: Gamma converting monitor ICC profile to printer's profile

Post by jsachs »

You can't determine this from the profile - it depends on the printer and the printer driver as well as the profile since they all work together as a system. The only way to determine the gamma is to print a test chart and read it with a densitometer.
Jonathan Sachs
Digital Light & Color
tomczak
Posts: 1367
Joined: April 25th, 2009, 12:56 am
What is the make/model of your primary camera?: Fuji X-E2
Contact:

Re: Gamma converting monitor ICC profile to printer's profile

Post by tomczak »

Many thanks.

There are several nice and clever visual tests that can closely estimate (or at least check) the system gamma for a monitor. They won't work for prints. Is there anything similar for printers (i.e. a test image that could be used to estimate a gamma of a printing system with a reasonable accuracy - all I'm doing now is looking at the black-to white ramps and eyeballing if they look any close to how they should be)? Cheers.
Maciej Tomczak
Phototramp.com
jsachs
Posts: 4203
Joined: January 22nd, 2009, 11:03 pm

Re: Gamma converting monitor ICC profile to printer's profile

Post by jsachs »

If you purchase an inexpensive Kodak Q-13 step wedge and print the simulated version you can download from our web site, you should be able to compare them, but this will not take into effect differing DMAX and DMIN values.
Jonathan Sachs
Digital Light & Color
tomczak
Posts: 1367
Joined: April 25th, 2009, 12:56 am
What is the make/model of your primary camera?: Fuji X-E2
Contact:

Re: Gamma converting monitor ICC profile to printer's profile

Post by tomczak »

I don't have a densitometer, but have a camera with a spot meter. Curiosity took the better of me I figured that maybe I could wing it. Here is what I've done:

I've printed 5, solid gray, pretty large squares. The pixel values for those squares in 8 bit were: 0, 64, 128, 192, 255. Then I spot-metered them under the same iluminance, noting ISO, shutter speed and aperture that the camera wanted to use to make them all look medium gray (that's how spotmeters are supposed to be calibrated). The idea is that knowing those three values, one could estimate the luminance of the square, and since the iluminance was constant, luminance will depend only on the percived reflectivity (density) of the square. It's an approximation, but the errors are likely systematic, and I'm interested in estimating the shape of the gamma curve of the printer system, so that may work.

The values for luminance (Cd/m^2) that I calculated for those squares (from black to white) were: 23.5, 59, 176, 353, 701. I measured it in relatively bright shaded daylight at some 3000Lux. I did it a few times: the precision was there, the absolute accuracy of the estimates - I have no idea... hopefully not that far off.

If both the luminance and the pixel values are scaled by white to a fraction of 0 to 1 range, I believe the following printer gamma transfer function could be applied:

Print_Luminance=Print_Black_Point_Luminance+Pixel_Value^gamma

When I fitted this equation to the 5 measured points, it fitted nicely with the gamma around 2.43 (L=0.0335*Value^2.43). I kind of expected high gamma, since the entire reason why I started doing all that was that I believed that the printer gamma was way too high - the shaddows turning out too dense and blocked, even under high illuminance.

But another thing that can be calculated from the white and black squares luminance values is Dmin and Dmax. Dmin seems to be around 0.1 and Dmax around 1.6...?! While this is consistent with my gut feeling while looking at the prints (some 1:30, 1:40 contrast ratio), and the measurement method is far from perfect, viewing conditions far from laboratory ideal, flare and stray light present etc., but that's how we see the actual prints, isn't it?, but these are Drurst Lambda prints on Kodak Endura Pro paper (I think the advertised value for Dmax is 2.3). What's going on here?

Can someone tell me if I'm making any sense?
Maciej Tomczak
Phototramp.com
jsachs
Posts: 4203
Joined: January 22nd, 2009, 11:03 pm

Re: Gamma converting monitor ICC profile to printer's profile

Post by jsachs »

If you mail me a test print I should be able to measure it with my reflection densitometer - haven't used it in a while but if I dust it off it might just work. A camera light meter is not really a very accurate way to measure a target.

My address is:

Jonathan Sachs
7 Gregory Rd.
Lyme, NH 03768
Jonathan Sachs
Digital Light & Color
tomczak
Posts: 1367
Joined: April 25th, 2009, 12:56 am
What is the make/model of your primary camera?: Fuji X-E2
Contact:

Re: Gamma converting monitor ICC profile to printer's profile

Post by tomczak »

Many thanks Jonathan - I'll send them tomorrow. But is Dmin~0.1 and Dmax~1.6 for Durst Lambda on Endura Pro matte paper is realistic at all?

I'm sorry for dragging this topic, but I was hoping to learn from people that are way more experienced in proper printing than I am. And it's also more or less B&W printing problem, eventhough done in colour, which I know little about.

One could test the printer with test images - say Q13 step wedge, but this is not that high of a bar for a printer to pass - the difference between two steps takes several pixel levels in 8 bit.

At the other end of the spectrum, there are very demanding test images (mostly for testing monitors), that for instace check if you can distinguish between all 0-9 and 246-255 values at the end black and white tails. Prints can't differentiate between them all...

Given that it is said that we could distinguish not more than 200 shades of gray under ideal conditions, and that the prints can have a contrast ratio is 1:200 at the very most, normally much less, especially when some flare is present, what is a fair objective contrast test that the print could be expected to pass (i.e. what level of pixel levels blocking would be considered normal, or good)?

The reason for the question is that it looks to me that some of the Durst Lambda Prints either truncate or compress to undetectibility some 10% of pixel levels in deep shaddows, either due to truncating black point or too high printing gamma.

One can say: test print and choose what looks good, but then sometimes you can't do that - I need to print remotely sometimes and have to rely on colour-management and lab quality and it's printer calibration, and other people's eyes. Softproofing is supposed to try to do it, but I can rarely get it to work properly. And I'm not even that interested in colour constancy (I am, but that's a secondary concern) - I just want to get the contrast under control: I see things on the screen, and then they are gone on the print, with everything supposedly callibrated, profiled etc.

If I knew what to expect from a high quality, high Dmax prints (Durst Lambdas and Light Jets are supposed to be this way), I could pre-prepare the prints better to show what I want them to show.

What do others know? and how to deal with it? Cheers.
Maciej Tomczak
Phototramp.com
JML
Posts: 44
Joined: April 25th, 2009, 9:47 am

Re: Gamma converting monitor ICC profile to printer's profile

Post by JML »

I don't print much but why not ask the Durst Lambda people what grayscale levels at the extremes they should be able to print and distinguish on the paper of their choice. Then make test prints with dark and light blocks on that paper and see what the printer you are using actually delivers? If it is sub-par, take up the issue with the vendor.
Post Reply