Bug with Working Space Tone Response Curve and bundled profiles

The sensor ADC, wether its in a film scanner or a camera, spits out linear light intensity equivalent codevalues usually. The problem for a loooong time was the amount of data created from this is insane. By switching to a log encoding you could store 10bit logarithmic codevalues and reproject onto 12 to 14 bit linear. Of course you had losses in precision, but that was creating less visible artifacts after the whole toolchain than storing 10bit linear. Each stop of light gets the same number of codevalues to be represented by. This goes nicely with human perception (roughly at least, with limits, but still).

There are many many more reasons for doing something like this, most guided by the idea that data storage is limited and expensive and compute time is expensive. Uncompressed raw at 2K resolution for a 2h film deliverable? Film shooting ratio is 10:1 on a exceptionally good shoot, meaning for a 2h movie there is 20h of raw footage at least. 2 camera angles? shitty actors? Soon you are at shooting ratios of 50:1 ! Negative film is expensive, sure, but uncompressed raw moving pictures? With backups? shooting 10bit log in a 4:2:2 (or even 4:2:0 !) color undersampled format at 2K resolution was (is) good enough for many many projects.
(today even ‘8bit log cameras’ have an image that is (just barely, but still) good enough for documentary work and the occasional film. shocking to some in the world of raw photography.)

Partly true and also not the whole story. Standards exist for (sometimes mediocre) reasons and at some point people agreed on that 10bit log format as a quasi-raw because real RAW was not practical, for storage. Whether you do your math in linear or log is not really that important as long as the math is right. Whatever gets projected is linear light and color again (there is your WYSIWYG maybe), what happens in between is a matter of taste if you will, not as elegant, sure but economical. Pretending that people in the film industry did not think about stuff is a bit of an underestimation.

:smiley: Sounds like a cult I want to join! Who is volunteering to be the first sacrifice? I envision a bloody ritual at moonlight for the greater good!

Henri Cartier-Bresson did not know how to raytrace and optimize a lens construction for the leica camera he used, yet he was a master at using this tool. That superficial hands-on survival instinct might be good enough for making art, at least sometimes, maybe?

2 Likes

You’re speaking for yourself?
https://discuss.pixls.us/t/darktable-3-0-for-dummies-hardcore-edition/15864

There is a trick I took from theACES

wait for someone to fact-check and confess.

Take a look at my discussion and look at the date
https://discuss.pixls.us/t/out-of-gamut-colors-matrix-hsv/17165/2

No. No. And no again. Do you even understand what HSL is ? It’s bloody RGB 3D space turned into polar coordinates.
It sucks. It deeply sucks.

Now look what people involved with aces really think about HSV

https://community.acescentral.com/t/gamut-mapping-in-cylindrical-and-conic-spaces/2870

However it’s nothing new, Adobe use hsv from the 90’s for gamut compression.

You really should be more open to discussion

Bob, thanks for the considered answer. I’d heard various of what you related before, was good of you to stitch it together for me.

Suboptimal choices are the bread and butter of engineering. I think the pragmatic philosopher Mick Jagger put it well: “You can’t always get what you want… but if you try sometime you find,
you get what you need.” :stuck_out_tongue:

1 Like