How to create a CALIBRATED albedo texture?

tmammela

Hi there,

I'm struggling to understand how calibrated albedo maps are actually made. There is so much talk about PBR, how it's wonderful, physically correct awesomeness. But not even Allegorithmic has any information on how to make a calibrated texture.

Let's look at an example. Pretend that I want to know what is the physically correct value for the blue color in my notebook. So I take this photo with my raw-capable camera. Lighting should be the same for the chart and the notebook, not too much specular or anything going on.



If I color pick the notebook-blue in Photoshop from this ugly straight from the camera photo, I get sRGB: 3, 149, 217.

Of course that is pointless, so let's make this a calibrated photo. I follow the guidance from datacolor calibration guide: http://cdn.spyder.datacolor.eu/wp-content/uploads/2014/08/SpyderCheckr_24_UserGuide_EN_v1.pdf?iv=23

They tell me to set white balance from grey patch E2, and then adjust E1 to 90 % and E6 to 4%. They tell me to use "exposure and black/shadows adjustment". So I do exposure -0,30 and blacks -0,20. Here is the result:



Next I do the whole calibration thing just like the guide tells me to. I get this as a result:



Okay great, something is being adjusted clearly. So now this should be calibrated photo? I will pick the blue from my notebook once again and I get sRGB: 95, 145, 199. Well that is a huuuge difference, I should be happy and paint my PBR thingy with that albedo color? Not so fast.

http://cdn.spyder.datacolor.eu/wp-content/uploads/2014/08/SpyderCheckr-Color-Data.pdf?iv=23



First three numbers: LAB values, then sRGB, then AdobeRGB. This is the datasheet for the color chart, I must assume datacolor has used some million dollar equipment to determine true values for these colors, so if my calibrated photo matches these, it is correct and I can use it to build a true PBR texture.

So, gray patch 2E should be sRGB 202, 198, 195. I measure: 203, 204, 204. Not perfect but I guess it's ok.

5E gray should be: 80, 80, 78, it is: 58, 57, 56. Not good.

1F primary cyan should be: 0, 127, 159 but it is: 0, 149, 182. Not good.

5G apple green should be: 157, 188, 54 but it is: 184, 194, 65. Not good.

I'm starting to feel that this calibrated photo is far from calibrated. What do I do wrong? How do big companies actually take perfect albedo values?

ILYHugh

Hey @tmammela A lot of how you color calibrate depends on the exposure you take with your camera, even with the calibrated chart, you can still be off by a few values for having underexposed and overexposed photos. Its good to try and white balance from the chart if it has a white and gray card during photo capture. The other hard part is if you don't have a color calibrated monitor it would be hard to edit without knowing what the value looks like. Big companies still have the same hard time trying to get the actual value, so you're not alone  Here is a link used an xrite color checker, the the charts are a bit different, but the premise should be the same!

marks

Tip: you can look up what the values of those patches are (eg what the sRGB color value should be for each color patch) - they're on wikipedia. Draw a big splodge over the top of each patch on a new layer in photoshop and you can easily see how close you are to what the value should be for each color patch.

tmammela

ILYHugh said:

Hey @tmammela A lot of how you color calibrate depends on the exposure you take with your camera, even with the calibrated chart, you can still be off by a few values for having underexposed and overexposed photos. Its good to try and white balance from the chart if it has a white and gray card during photo capture. The other hard part is if you don't have a color calibrated monitor it would be hard to edit without knowing what the value looks like. Big companies still have the same hard time trying to get the actual value, so you're not alone  Here is a link used an xrite color checker, the the charts are a bit different, but the premise should be the same!

As you can see the color chart includes gray patches which are used for white balance in Lightroom. What do you mean "white balance during photo capture"?

Why would I need a calibrated monitor? I'm not adjusting anything by eye. There should be a mathematical way to do this correctly, or else it wouldn't be physically based. Did you read my post? I'm not judging the colors by eye. I'm using values measured from the photo and comparing those to reference values.

marks said:

Tip: you can look up what the values of those patches are (eg what the sRGB color value should be for each color patch) - they're on wikipedia. Draw a big splodge over the top of each patch on a new layer in photoshop and you can easily see how close you are to what the value should be for each color patch.

How does Wikipedia values differ from that chart I posted which is made by the same company which makes the actual color chart?'

EDIT: ILYHugh thanks for the link, there was some useful stuff there. 3D Lut Creator seems interesting, I tried the demo and found out that reference values for Spyder Checkr24 are different there compared to the official datasheet. I emailed them about this, curious to know why the values are so different.

Also there was one obvious mistake I made with the test images in opening post. They were taken in LED lighting. Stupid! LED's have such bad color rendering index it's only natural colors will be wildly different than in daylight. BUT. I haven't been able to reproduce correct values from daylight shots either, yet.

radiancef0rge

Did your color checker come with software? An xrite chart typically comes with software that corrects per channel to match all the squares at once as opposed to eyeballing an adjustment until the value is matching with a color picker. 

leleuxart

You can try DaVinci Resolve for this as well. They have a selection of charts to match against, plus options for white balance and color temperature. You'll have to know what gamma space you shot in and what you'll be viewing in, as well as the target color space.