Kodak Q-13 color target

ggc · August 20, 2018, 12:58pm

Meanwhile, 0-100 values for colors.

|Name|R|G|B||R|G|B|
|---|---|---|---|---|---|---|---|
|||||||||
|Light_Blue|243473|240670|229074||95,48|94,38|89,833|
|Blue|113479|100212|151237||44,502|39,299|59,309|
|Light_Cyan|232292|237325|222258||91,095|93,069|87,16|
|Cyan|39255|157842|200724||15,394|61,899|78,715|
|Light_Green|222769|227829|192470||87,36|89,345|75,478|
|Green|83448|156321|78440||32,725|61,302|30,761|
|Light_Yellow|246280|236952|163699||96,58|92,922|64,196|
|Yellow|248129|222480|52150||97,305|87,247|20,451|
|Light_Red|234436|201450|155827||91,936|79|61,109|
|Red|191807|67359|70052||75,218|26,415|27,471|
|Light_Magenta|242981|222436|193704||95,287|87,23|75,962|
|Magenta|186805|24702|90269||73,257|9,687|35,4|
|White|247637|245723|237539||97,113|96,362|93,153|
|Light_Brown|195278|184085|132714||76,58|72,19|52,045|
|Brown|106988|89230|80456||41,956|34,992|31,551|
|Light_Gray|233699|231318|215430||91,647|90,713|84,482|
|Black|53318|50971|50392||20,909|19,989|19,762|

ggc · August 20, 2018, 1:02pm

Using rgb curves seems to create too strong discontinuitess, already with the first 4 boxes…

age · August 20, 2018, 1:53pm

Very informative!
From what I understand it says that Adobe apply some hidden exposure compensation at different iso and it shows how to neutralize that with the Kodak grey card, the most important patch is M because it is the middle gray 18% in linear gamma( 45 in 8 bit, 11654 in 16 bit, 116 in 8 bit and 2.2 gamma, and so on)

Morgan_Hardwood · August 20, 2018, 2:27pm

Yes, as do most raw converters. They apply a baseline exposure value and use a non-linear tone-curve.

To see the “unmodified” raw image, just open a raw file in RawTherapee and apply the “Neutral” profile.

Morgan_Hardwood · August 20, 2018, 2:39pm

I recorded this screencast to show how to use the HH curve (hue according to hue) to adjust the colors in the image to match some reference values:

ggc · August 20, 2018, 3:11pm

Thank you @Morgan_Hardwood .

I’ve tried, but seems to me it’s impossible to reach the suggested values…

E.g. in the upper left box, i should reach 243 240 229

LaurenceLumi · August 20, 2018, 3:20pm

Technically no, as I am pretty sure those guides are created with CMY dyes, and are not true colors. Its probably a piece of RA4 photographic paper. But I am sure that it works well enough. The Q60 and the IT8 ones are also made in this way, for the specific purpose of profiling the dyes. But they probably still work well enough as well.

The color checker, Macbeth etc are specifically designed for the job at hand.

Morgan_Hardwood · August 20, 2018, 3:22pm

@ggc 243 in what color representation (HSV, RGB, etc) and which color space (AdobeRGB, sRGB, etc)?
A brief search yielded various conflicting values reported in various forums but nothing that looked authoritative, so also beware of incorrect reference values.

ggc · August 20, 2018, 3:29pm

243 r, 240 g, 229 b, in adobe rgb color space.

Anyway i’m trying another way.

I have white balanced the image and i’m filling a gimp color chart i’m creating with sampled values.

Edit: then i’ll try to make them as the Fuji X-T20 one.

ggc · August 20, 2018, 3:45pm

I’ve filled all the cells.

But… should i use, in raw therapee to pick colors, no profile or camera standard? Makes a big difference.

camera standard:
2018-08-20-174416_1680x1050_scrot

no profile:
2018-08-20-174420_1680x1050_scrot

ggc · August 20, 2018, 3:48pm

The color table.

ggc · August 20, 2018, 4:03pm

Meanwhile, i attach the gimp .xcf file with the rulers and the fuji sampled points screenshot always wb adjusted.

ColorChart-Honor6A-wbadjusted.xcf.bz2 (4.1 KB)

sguyader · August 20, 2018, 4:22pm

@ggc if you have trouble tweaking the colors with the H=f(H) curve in Lab* Tool, you can try to do it in the LHS Equalizer (in the color tab). It has a H=f(H) curve which has more lattitude.

ggc · August 20, 2018, 5:09pm

Thank you. Now i’m trying with the g’mic way.

ggc · August 20, 2018, 5:10pm

The 2 color charts… unfortunatly the Fuji reaches 255 in a pair of boxes for 1 channel.

.

Now i need to update g’mic.

ggc · August 20, 2018, 5:26pm

I have generated the lut, and tried. Unfortunatly the result is really 8 bit like, posterized…

sguyader · August 20, 2018, 5:56pm

If what you want is make you phone camera’s pictures match roughly those of your XT-20, what i’d do is take a picture of the chart with both cameras.
In RT, open the XT-20 image and in color management, check that the DCP profile is loaded (automatic matching). Now for each patch, make L*, a* and b* measurements. For the grey patches, don’t measure all patches to avoid weird curve shape.
Now load the phone camera image, and create a luminosity tone curve to match roughly the grey patches L* values, avoiding any weird shape.
Then use the H=f(H), L=f(H) and C=f(H) curves in LHS Equalizer to match roughly a* and b* values while keeping the eye on the L* values. You may need to go back and forth between curves as tweaking one may alter the other.
It’s a long process but it should work reasonably well. I’ve used this technique to produce some “Fuji simulation presets” and the results were pretty good.

ggc · August 20, 2018, 9:27pm

Thank you again.

For today i give up, and i think even for tomorrow.

Even if it’s far from being ideal, what happens if i take a picture of the macbeth on the monitor with both the fuji and the phone, and use that? (maybe you have already tried…)

Thank you again.

ggc · August 22, 2018, 1:48pm

For now i’ve been able to take again the pictures, and i hope they will be ok this time

https://we.tl/t-RgDDXzhcqZ

ggc · August 23, 2018, 2:33pm

Here are reference values, in the shot made with the Fuji X-T20. White balanced on M.