HDR heatmaps and torch mode UHDs

[Pyoko]

Originally Posted by Geoff D

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

Yes, it all depends on the direction of the conversion error. It's like video levels where there have been both crushed discs incorrectly encoded with PC levels, and washed-out discs incorrectly encoded with "double" video levels.

[ghostly]

Originally Posted by Pyoko

Ghost in the Shell (1995) - Something a bit different this time. Instead of comparing the US UHD to the BD, the second set is the JP UHD. ...

What software(s) are you using for UHD image processing (couldn't find info in a quick scan of this thread)? Thanks.

[Pyoko]

Originally Posted by ghostly

What software(s) are you using for UHD image processing (couldn't find info in a quick scan of this thread)? Thanks.

I'm using Nuke, a compositing application.

[Agent Kay]

New mutants I bet

[Pyoko]

Shots are from Episode 1 and 2.

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 102 nits (Gamma at 2.30), based on 20 samples. (Min: 97, Max: 104, Median: 103)

[Scottishguy]

[Show spoiler]

Originally Posted by Pyoko

Shots are from Episode 1 and 2.

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

https://youtu.be/r7hEsKT2e_c

[Fjodor2000]

Originally Posted by Pyoko

Shots are from Episode 1 and 2. [Chernobyl]

There was a similar analysis made by BrownianMotion previously here. In that analysis it looked like there were quite a lot of colors close to the DCI-P3 triangle border, which means outside of Rec.709, and thus WCG.

So I would have expected that to also show up more distinctively in your WCG analysis too. But perhaps the reason for the differences simply is that you have selected different sample images to analyze, and the amount of WCG varies a lot throughout the movie?

I guess this is also a general question -- i.e. what methodology do you use when deciding what image samples to analyze? E.g. are you selecting them based on if they are sticking out in some regard (e.g. high max brightness), or being representative for the movie in general, or just taking a sample every e.g. 20 minutes?

[Pyoko]

Originally Posted by Fjodor2000

There was a similar analysis made by BrownianMotion previously here. In that analysis it looked like there were quite a lot of colors close to the DCI-P3 triangle border, which means outside of Rec.709, and thus WCG...

Good questions. The sample choice is obviously a deciding factor, but as I explain in the first post my color gamut analysis also attenuates the values according to the luminance (and in the chart somewhat depending on the number of pixels with that value, though I've been thinking of tweaking this) otherwise I end up with too many artifacts, and I think it's a misrepresentation to clearly show saturated values from areas that are too dark to appreciate anyway, or random pixels that are just color noise. BrownianMotion and I happened to choose frames from the same shot if you want to compare. Here's what it would look like for me without any attenuation:


Also keep in mind that P3 and 709 share the blue primary, so values around that corner in P3 will still be mostly the same in 709.

As for the selection I focus on choosing samples that would be interesting and actually show something, which generally means shots with light sources or stronger reflections or colors, otherwise there would be no point. As such these analyses are not good representations of how much HDR there is in any given show, but representations of how strong it is on those shots that have any HDR potential to begin with.

WCG is much harder to judge if it'll show up or not. Sometimes you take a capture with a colored light or something similar in it that looks like it would for sure give a reading beyond 709, but it doesn't. In the case of Chernobyl these are the two most colorful samples I happened upon:

[Fjodor2000]

Originally Posted by Pyoko

As for the selection I focus on choosing samples that would be interesting and actually show something, which generally means shots with light sources or stronger reflections or colors, otherwise there would be no point. As such these analyses are not good representations of how much HDR there is in any given show, but representations of how strong it is on those shots that have any HDR potential to begin with.

An interesting analysis would be to see how many WCG or HDR pixels there are per frame on average in the complete movie. Or the average percentage of such pixels. That would give an estimate of the overall/average WCG or HDR in the movie.

To do such an analysis, one would have to extract all frames, and count the number of WCG/HDR pixels in each frame. Sounds like a lot of work. And it would be, for a human. But for a computer it's definitively doable. As a comparison a transcode of a 4K movie takes a couple of hours on a powerful modern CPU, and that work also extracts all frames and does calculations for each pixel in each frame. In addition, the computation done per pixel is much heavier than just counting if it's WCG/HDR or not.

So such a "WCG/HDR analysis program" should be able to automatically process a 4K movie within hours too.

Might be an interesting hobby project for someone with spare time during the Christmas Holidays or whenever.

[andreasy969]

I'd say it'd be mostly interesting for those who care about numbers. But since one has a hard time telling/guessing where there actually is WCG in the first place, the practical value is rather low IMHO. If anything, I'm sure we'd get a very low pixel percentage with most movies. I'd also love to see certain folk reliably pointing to WCG shots (without lamps).

[Fjodor2000]

Originally Posted by andreasy969

But since one has a hard time telling/guessing where there actually is WCG in the first place, the practical value is rather low IMHO.

Wouldn't it be the opposite then, i.e. it would highlight the amount of WCG pixels that could be hard to determine otherwise?

Or do you mean that it's irrelevant anyway, since WCG does not matter for image quality?

[andreasy969]

Originally Posted by Fjodor2000

Wouldn't it be the opposite then, i.e. it would highlight the amount of WCG pixels that could be hard to determine otherwise?

Yes, but my point was why bother with sth that one cannot really see? As in "There's clearly expanded gamut in this shot!". Most likely one will just prove that the expanded gamut is very overrated, but one can see that already. And even if other gamut graphs show relatively large areas, that might still be rather isolated spots.

Or do you mean that it's irrelevant anyway, since WCG does not matter for image quality?

That'd be too harsh, but yes, even though it being extremely overrated is more like it. I mean just look at all the b&w caps.

I guess my main point is that Pyoko's gamut caps show all one needs to know already IMO.

[Fjodor2000]

Originally Posted by andreasy969

That'd be too harsh, but yes, even though it being extremely overrated is more like it. I mean just look at all the b&w caps.

I guess my main point is that Pyoko's gamut caps show all one needs to know already IMO.

I agree it's an interesting analysis. But the problem it just shows the amount of WCG for one single frame (or a few selected frames). It does not say anything about the average amount of WCG in the movie.

And in the end, you have to make up your mind. Does WCG matter for image quality or not? If it doesn't then any WCG analysis at all is pointless.

Also, all of the above applies to HDR too by the way.

[andreasy969]

Originally Posted by Fjodor2000

I agree it's an interesting analysis. But the problem it just shows the amount of WCG for one single frame (or a few selected frames). It does not say anything about the average amount of WCG in the movie.

And in the end, you have to make up your mind. Does WCG matter for image quality or not? If it doesn't then any WCG analysis at all is pointless.

Also, all of the above applies to HDR too by the way.

What random or even not so random frames show gamut wise, will be the same with the movie. Do you expect a face suddenly having more gamut in other frames for ex? The gamut is mostly artificial light sources.

I also don't/didn't say it doesn't matter at all. What I do say is that measuring the whole movie is pointless/not worth it IMO. It will only tell you exactly what you're seeing with the caps already: Expanded gamut will (with most movies) be a very low percentage.

And no, the same does not apply to HDR: The HDR is much more obvious and one can easily tell without having to look at a graph - that's both re. brightness and detail.

[Fjodor2000]

Originally Posted by andreasy969

What random or even not so random frames show gamut wise, will be the same with the movie. Do you expect a face suddenly having more gamut in other frames for ex? The gamut is mostly artificial light sources.

I expect the amount of WCG to vary through the movie. Hence the need for an analysis of average WCG. Single (or selected) frame analysis will be interesting, but not representative of the overall WCG. The same goes for HDR.

Originally Posted by andreasy969

I also don't/didn't say it doesn't matter at all. What I do say is that measuring the whole movie is pointless/not worth it IMO. It will only tell you exactly what you're seeing with the caps already: Expanded gamut will (with most movies) be a very low percentage.

Point taken, and I agree with the part in bold.

E.g. for some movies 4K resolution or HDR may improve image quality more than WCG. But this will differ per 4K BD movie. So the amount that 4K vs HDR vs WCG improves the image quality compared to the BD will differ per movie. And all three are interesting to analyze.

Originally Posted by andreasy969

And no, the same does not apply to HDR: The HDR is much more obvious and one can easily tell without having to look at a graph - that's both re. brightness and detail.

You mean that for 4K BD, HDR is more important for image quality than 4K or WCG, regardless of movie, and regardless or corresponding BD?

[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 123 nits (Gamma at 2.40), based on 20 samples. (Min: 83, Max: 147, Median: 121)

[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 86 nits (Gamma at 2.30), based on 20 samples. (Min: 65, Max: 127, Median: 81)

[Fjodor2000]

For the last two analyses (Sherlock Holmes and Collateral), there is no color in the WCG analysis images showing colors outside of Rec.709 gamut. I.e. for some images the "gamut triangles" show colors outside Rec.709, but the image itself is still completely black&white.

Is this intentional, so you have change they way it is presented?

[Pyoko]

Originally Posted by Fjodor2000

For the last two analyses (Sherlock Holmes and Collateral), there is no color in the WCG analysis images showing colors outside of Rec.709 gamut. I.e. for some images the "gamut triangles" show colors outside Rec.709, but the image itself is still completely black&white.

Is this intentional, so you have change they way it is presented?

I messed around with it a bit but I think I reverted it to the way it was before in the end. If there are some subtle or stray readings outside 709 in the gamut chart but it's not really visible in the image then it's almost certainly just noise that's not really appreciable anyway. In the second to last shot for example, if you look closely, you can see some slightly orange pixels here and there.

[Fjodor2000]

Originally Posted by Pyoko

I messed around with it a bit but I think I reverted it to the way it was before in the end. If there are some subtle or stray readings outside 709 in the gamut chart but it's not really visible in the image then it's almost certainly just noise that's not really appreciable anyway. In the second to last shot for example, if you look closely, you can see some slightly orange pixels here and there.

Hmmm... I'm not seeing anything really, looks all B&W to me. Haven't pixel peeped though.

But does it mean that these two movies are more or less "standard color gamut (Rec.709) in WCG container" then (at least for those screenshots that have been analyzed)?