2021-03-28, 05:17 PM
Okay I got a new candidate. V2 Variant 3
framecompare decided it doesnt like me anymore, so now I'm posting it on slow.pics
https://slow.pics/c/zLZeZmKg
The order is counterintuitive - Variant 3 comes as the second image, because it's the closest to the look of Variant 1.
It does - for now - not incorporate the DCI-P3 idea because I've changed the entire workflow and I'm not sure how to incorporate it yet (if at all).
Description of what I did: First off, I stopped relying on any Adobe color conversions. Something is weird with them I think. Instead it's all manual via matrices now. Also, this time I did not use a fixed conversion between predefined white points. Instead I simply measured the white point in correctly (?) converted XYZ, then created a conversion between that white point and D65. Then converted back to Rec 2020, and voila, I think I hit the white right in the head.
So this should, in theory (well, I've been wrong enough times before), be a pretty scientifically accurate white balance correction to the white point (determined by measuring the average values in a blurred B&W Justice League).
The determined white point's x,y coordinates appear to be: 0.325, 0.3463.
A quick look at https://en.wikipedia.org/wiki/Standard_i...lluminants tells us that this isn't really anything standard (at least nothing on that list), however it's "reasonably" close to D55.
What does this mean? No clue. Was it simply the display they mastered on? Did they make some weird color conversions that turned it into something weird?
Given that this seems to relatively consistently hit the white on stuff like titles and also the B&W version of the movie (which, well, should be monochrome?), I can't imagine that it's just a creative decision, otherwise you'd expect it to vary from scene to scene.
I still have the suspicion that the saturation might be a tad lower than intended, so maybe there was indeed some quirk in converting between DCI-P3 and Rec709 or something like that...
Okay, long story short, I'm completely clueless and this is basically just a glorified scientifically accurate (?) version of "muh curves".
TLDR: It's pretty similar to V1 but just kinda looks a tad more harmonic and not a tiny tad too blue as V1 did.
I'm not settling on anything just yet and I think I'm gonna keep this project up as a long-term thing where I try new ideas wherever I learn something new until maybe one day, I find the true, absolute and undeniable answer. Maybe this Variant 3 is already the correct answer ... it just feels very unsatisfying to me because I invested so much time trying to find a hidden pattern only to find ... some random white point that can be white balanced. Doesn't feel right, man!
Edit: On the other hand, maybe the reason I thought it didn't look properly saturated, especially in the reds, is because it was a tad too blue, suppressing the reds. They do look more natural in Variant 3 now..
framecompare decided it doesnt like me anymore, so now I'm posting it on slow.pics
https://slow.pics/c/zLZeZmKg
The order is counterintuitive - Variant 3 comes as the second image, because it's the closest to the look of Variant 1.
It does - for now - not incorporate the DCI-P3 idea because I've changed the entire workflow and I'm not sure how to incorporate it yet (if at all).
Description of what I did: First off, I stopped relying on any Adobe color conversions. Something is weird with them I think. Instead it's all manual via matrices now. Also, this time I did not use a fixed conversion between predefined white points. Instead I simply measured the white point in correctly (?) converted XYZ, then created a conversion between that white point and D65. Then converted back to Rec 2020, and voila, I think I hit the white right in the head.
So this should, in theory (well, I've been wrong enough times before), be a pretty scientifically accurate white balance correction to the white point (determined by measuring the average values in a blurred B&W Justice League).
The determined white point's x,y coordinates appear to be: 0.325, 0.3463.
A quick look at https://en.wikipedia.org/wiki/Standard_i...lluminants tells us that this isn't really anything standard (at least nothing on that list), however it's "reasonably" close to D55.
What does this mean? No clue. Was it simply the display they mastered on? Did they make some weird color conversions that turned it into something weird?
Given that this seems to relatively consistently hit the white on stuff like titles and also the B&W version of the movie (which, well, should be monochrome?), I can't imagine that it's just a creative decision, otherwise you'd expect it to vary from scene to scene.
I still have the suspicion that the saturation might be a tad lower than intended, so maybe there was indeed some quirk in converting between DCI-P3 and Rec709 or something like that...
Okay, long story short, I'm completely clueless and this is basically just a glorified scientifically accurate (?) version of "muh curves".
TLDR: It's pretty similar to V1 but just kinda looks a tad more harmonic and not a tiny tad too blue as V1 did.
I'm not settling on anything just yet and I think I'm gonna keep this project up as a long-term thing where I try new ideas wherever I learn something new until maybe one day, I find the true, absolute and undeniable answer. Maybe this Variant 3 is already the correct answer ... it just feels very unsatisfying to me because I invested so much time trying to find a hidden pattern only to find ... some random white point that can be white balanced. Doesn't feel right, man!
Edit: On the other hand, maybe the reason I thought it didn't look properly saturated, especially in the reds, is because it was a tad too blue, suppressing the reds. They do look more natural in Variant 3 now..
![[Image: G4ccTRw6_o.jpg]](https://images2.imgbox.com/30/dd/G4ccTRw6_o.jpg)
