HDR heatmaps and torch mode UHDs

[Christian Muth]

Originally Posted by andreasy969

Fair enough. And as I said: It looks very good.


This one - the "torch mode" thread.


I cancelled my Synapse UHD-BD comparison myself. I started it, but aborted it due to requiring too much time at the time.

Gottcha! Otherwise, I would've posted this:

Them's Fightin' Words - Simpsons Clip - YouTube

But you seem reasonable

Chris

[mrtickleuk]

Originally Posted by tama

I believe they're the same thing. In 2017 LG added it but to activate it you set dynamic contrast to low. It didn't have its own function in the menu.

In 2018 LG created its own function as Dynamic Tone Mapping in the TV's menu.

I just referred to it on my TV the way LG had referred to it for my model.

Cool, that explains it. Thanks!

[Pyoko]

Originally Posted by Christian Muth

I have a question re: "forced DV" (which may have already been answered so please don't rip me a new one for asking again)- I have a Sony UBP-X700 set to Dolby Vision output with my LG 65B7A, so everything is essentially "forced DV", and does this actually "change" the playback quality of HDR-10 UHDs, or is it simply outputting the HDR-10 of those discs in a "Dolby Vision wrapper", so to speak? (ducks in case this has already been answered and folks here decide to rip me a new one for asking such a stupid question)...

It's just the same image wrapped in DV on the source side, but this forces it to go through the DV processing path on the TV which on the 7-series is better than the HDR10 processing path.

[Christian Muth]

Originally Posted by Pyoko

It's just the same image wrapped in DV on the source side, but this forces it to go through the DV processing path on the TV which on the 7-series is better than the HDR10 processing path.

Thank you, sir!

Chris

[Pyoko]

UHD color | UHD luminance | BD luminance | UHD tonemap | BD upscale

With a similar highlight roll-off the HDR version would on average be equivalent to the SDR version with peak white at 162 nits (Gamma at 2.40), based on 20 samples. (Min: 111, Max: 264, Median: 157)

[Scottishguy]

Well that confirms what I thought about Top Gun. Sensible and effective HDR. It's not trying to drastically alter the image, but complimenting it to enhance it's cinematic quality for home presentation.

[Pyoko]

Excellent sharpness and detail on this Alexa 65 footage. Very clean look, except perhaps for some of the lenses exhibiting a bit of chromatic aberration. The BD appears low-pass filtered so the upcoming Criterion one might look a bit better.

UHD color | UHD luminance | BD luminance | UHD tonemap | BD upscale

With a similar highlight roll-off the HDR version would on average be equivalent to the SDR version with peak white at 181 nits (Gamma at 2.40), based on 20 samples. (Min: 164, Max: 211, Median: 178)

[tama]

I'd love to see Sony's The Natural

@Pyoko
There is a SDR 4K version of "Parasite". This release was not intended to be SDR and they released a HDR version a bit later. I have both versions and can take screenshots if anyone is interested.

The interesting part ist that the saturation of both versions only fit if I play the SDR version interpreted as having DCI-P3 gamut. Otherwise the SDR version looks too undersaturated. Ah and the SDR version has 10 Bit, too.

With madVR's tone mapping bot version look very similar, but not exactly the same.

[Scottishguy]

Originally Posted by FoLLgoTT

@Pyoko
There is a SDR 4K version of "Parasite". This release was not intended to be SDR and they released a HDR version a bit later. I have both versions and can take screenshots if anyone is interested.

The interesting part ist that the saturation of both versions only fit if I play the SDR version interpreted as having DCI-P3 gamut. Otherwise the SDR version looks too undersaturated. Ah and the SDR version has 10 Bit, too.

With madVR's tone mapping bot version look very similar, but not exactly the same.

There's an interesting situation with The Limey that you have to force wide gamut to achieve proper colour saturation. I'm guessing cause it's down converted HDR into SDR.

[Fjodor2000]

It would be really nice if blu-ray.com would include analyses like the ones in this thread for all 4K UHD Blu-rays they review.

Assuming they would set up the "tools" once, the process could be automated, right?

[FoLLgoTT]

Strange. My account was deleted automatically. I guess, because of my signature. Ok, I'm here again.

[Scottishguy]

Originally Posted by FoLLgoTT

Strange. My account was deleted automatically. I guess, because of my signature. Ok, I'm here again.

Welcome back.

[Cactus]

@Pyoko: Would it be possible for you to share your workflow for these images? I would really like to try this myself.

[Pyoko]

Back from the dead because I was interested to see the numbers on this one (and also because I fixed my script which I inadvertently broke and had forgotten to back up.) Here I've linearly increased the brightness of the nits analysis of the SDR to roughly match the HDR, for a more apples to apples comparison of what the expansion on the HDR does.

UHD color | UHD luminance | BD luminance | UHD tonemap | BD upscale

With a similar highlight roll-off the HDR version would on average be equivalent to the SDR version with peak white at 173 nits (Gamma at 2.40), based on 20 samples. (Min: 124, Max: 234, Median: 175)

[Pyoko]

Originally Posted by Cactus

@Pyoko: Would it be possible for you to share your workflow for these images? I would really like to try this myself.

Here's a screenshot of my workflow:


Do you have experience with Nuke? I'm somewhat wary of sharing this stuff simply because it's about as far from fool-proof as you can get, so easy to mess something up if you don't know exactly what each thingamabob in the script does.

[Pyoko]

Although some shots show a peak upwards of 1000 nits it's usually just a couple of stray pixels somewhere, mostly it keeps to around 300 nits appropriately enough . Perhaps I should try to code some kind of histogram thingy.

UHD color | UHD luminance | BD luminance | UHD tonemap | BD upscale

With a similar highlight roll-off the HDR version would on average be equivalent to the SDR version with peak white at 101 nits (Gamma at 2.40), based on 20 samples. (Min: 88, Max: 118, Median: 100)

[K9 Mark V]

Scarface, The Dark Knight Trilogy, and Batman 89/Batman Returns please?

[Pyoko]

Something a bit different this time. Instead of comparing the US UHD to the BD, the second set is the JP UHD. Similarly to the previous couple of entries the luminance of the JP is linearly scaled/brightened to roughly match the US, and the numbers in parentheses on the JP images are the actual unscaled nit numbers in the stream.

Two major observations from this:

While the base target brightness of the US UHD is almost twice that of the JP (and in my opinion too bright), the US peaks reach 2-3 times even the scaled JP peaks so the upper end has been non-linearly expanded quite a bit to get that HDR effect.

We appear to once again have a mistake in the form of an incorrect DCI-P3<>Rec2020 conversion akin to Watchmen and Braveheart. Comparing the gamuts you can see they look rotated/skewed most of the time and the effect is mostly visible in blue-green tones, where the JP goes more towards cyan, while the US goes more towards blue. This is very clear on the aquarium in shot 3 or the glow around the sparks in shot 10.

So which one is correct? On the one hand the JP UHD does go out into Rec2020 territory at times, on the other looking at andreas's screenshot comparison to the old BD master, it's obvious the cyan tones of the JP match much better. It's not a huge deal here as both looks kind of work for the movie, but to my eyes the JP colors look more correct and more sensible overall, with the US looking more "off" when flipping between them, slightly desaturated with a kind of blue-purple wash.

It should be noted however that it is not a straight master conversion error like Watchmen. Here's the aquarium shot, first JP and then US. Clear difference with skewed gamut.


Next we have the JP vs the US with the DCI-P3<>Rec2020 conversion fix. The direction of the gamut is now aligned, but it ends up oversaturated.


Finally we have the JP vs the US with the fix, plus the saturation slightly scaled back. Now it matches pretty well.


What I suspect may have happened is they got the master from Japan, and upon doing their own HDR pass incorrectly assumed the wrong colorspace when importing it, got a skewed and desaturated starting point, and then had to increase the saturation a bit to compensate (but with the skewed colors remaining.)

UHD (US) color | UHD (JP) color | UHD (US) luminance | UHD (JP) luminance | UHD (US) tonemap | UHD (JP) tonemap

With a similar highlight roll-off the HDR (US) version would on average be equivalent to the HDR (JP) version with peak white at 175%, based on 20 samples. (Min: 131%, Max: 317%, Median: 170%)

[Geoff D]

Interesting that the colour space error makes it less cyan though, that's the exact opposite of your Bravehearts and your Watchmens.