AVS Forum banner

761 - 780 of 1220 Posts

·
Registered
Joined
·
173 Posts
I think why I noticed the difference was because in that infinity war scene many of the clipped frames ended up with peaks bellow the minimal target and so it was effecting the target nits.
 

·
Registered
Joined
·
420 Posts
App keeps re measuring files

Ok so the app re measures files previously measured for me. Im wondering if this is because I have the measurement files be created in a separate folder. Not in the movie folder.
Thoughts?
 

·
Registered
Joined
·
1,479 Posts
Discussion Starter #763
Ok so the app re measures files previously measured for me. Im wondering if this is because I have the measurement files be created in a separate folder. Not in the movie folder.
Thoughts?
Yes. The tool is not optimized for that.
Especially because the measurement file get renamed automatically by madmeasurehdr to include the path to the mkv in the name.

This makes things more difficult to work with.
If possible, keep your measurement next to your mkv. The tool works perfectly for that. ;-)
 

·
Premium Member
Joined
·
1,760 Posts
EDIT: after talking further to madshi, I realized we did not give him another part of the algo to avoid doing dynamic clipping when not worth it. It was included in the scene part of the algo that he did not need since he can only work frame based.

Right now, current version of the dynamic clipping in madVR is basically clipping every single frames at the knee.

In the dynamic clipping of our tool, we have introduced safeguard:
1) Work scene based an not frame based as discussed before
2) Do not clip if the decompressed area is not big enough in comparison to the clipped area.
If decompressed area < 10* clipping area, then we say it's not worth it, and we do not apply dynamic clipping.

This can be seen for example in the mother scene with the lamp ( @Neo-XP)
--> Current version of dynamic clipping in madVR is clipping part of the lamp
--> Our tool does not apply clipping there. It looks at the potential decompressed area (75nits to clipped peak), and says decompressed area is lower than 10 times the clipped area, so better not clip there.

I have proposed to madshi to not only look at the size/area of the decompressed pixels, but also to look at how strongly would they be decompressed as a further check.

In any cases, the checks 1) and 2) are sorting out A LOT of frames. For example, we only clip (more than 10nits) 30% of the frames of Lucy. Very likely, the number is quite a lot higher in the current madVR beta V41.
Does version 43 that madshi released today fix this?
 
  • Like
Reactions: chros73

·
Registered
Joined
·
1,479 Posts
Discussion Starter #765 (Edited)
Does version 43 that madshi released today fix this?
No. But we will soon release a new version ourselves to make optional to be scene based (current) or frame based (more aggressive) and a few light changes to the dynamic algo. :)

MadVR current version is still missing the "do not clip" if it's not worth it step. It can clip highlights even if there is no benefit/decompression in the picture from it at the moment.
 

·
Registered
Joined
·
7,973 Posts
As well as the FALL algo seems to work for projectors, there's one thing that confuses me:

Let's say I measure a movie with default settings and a minimum/real display peak of 100nits. Now let's say there's a really bright scene that with these settings gets assigned a dynamic target peak nits value of 1200nits. Now I change the settings and choose a minimum/real display peak of 1000nits, and run the optimization tool over it once more. Nothing changes. Still 1200nits dynamic target peak nits.

Does that really make sense? If we position two TVs next to each other, one with true 100nits peak nits, and one with true 1000nits peak nits. Can it really be correct that the optimization tool tells madVR to render the same scene 100% identically for both of these very different TVs?

Shouldn't the TV's real peak nits somehow affect the dynamic target peak nits value?
 

·
Premium Member
Joined
·
1,760 Posts
As well as the FALL algo seems to work for projectors, there's one thing that confuses me:

Let's say I measure a movie with default settings and a minimum/real display peak of 100nits. Now let's say there's a really bright scene that with these settings gets assigned a dynamic target peak nits value of 1200nits. Now I change the settings and choose a minimum/real display peak of 1000nits, and run the optimization tool over it once more. Nothing changes. Still 1200nits dynamic target peak nits.

Does that really make sense? If we position two TVs next to each other, one with true 100nits peak nits, and one with true 1000nits peak nits. Can it really be correct that the optimization tool tells madVR to render the same scene 100% identically for both of these very different TVs?

Shouldn't the TV's real peak nits somehow affect the dynamic target peak nits value?
I would think it should, otherwise how do you adapt the tone mapping to a specific display?
 

·
Registered
Joined
·
173 Posts
As well as the FALL algo seems to work for projectors, there's one thing that confuses me:

Let's say I measure a movie with default settings and a minimum/real display peak of 100nits. Now let's say there's a really bright scene that with these settings gets assigned a dynamic target peak nits value of 1200nits. Now I change the settings and choose a minimum/real display peak of 1000nits, and run the optimization tool over it once more. Nothing changes. Still 1200nits dynamic target peak nits.

Does that really make sense? If we position two TVs next to each other, one with true 100nits peak nits, and one with true 1000nits peak nits. Can it really be correct that the optimization tool tells madVR to render the same scene 100% identically for both of these very different TVs?

Shouldn't the TV's real peak nits somehow affect the dynamic target peak nits value?
Originally the dynamic tuning variable was the real display nits, but we found that it was very subjective and it made more sense to use that as the minimum target instead. So I think that someone with a brighter display would also use a higher dynamic tuning value. At which point maybe the 2x average highlight cap is going to be too low, so maybe at a certain real display or dynamic tuning value the average highlight cap should be dropped?

But also, even with the same target at 1500, the 1000 nits display isn't darkening the image as much as the 100 nits display, so it's going to look completely different.
 

·
Registered
Joined
·
1,479 Posts
Discussion Starter #770
@madshi
@Dexter Kane is right.

At the beginning, minimum/real target nits and dynamic tuning were one and the same.

That's why dynamic tuning goes from 50 to 200 recommended.

However, as experienced showed, people taste do not vary linearly with their real on screen brightness.

For example, @Neo-XP has about 150 real nits and we have 50 real nits. However both of us liked the same setting of dynamic tuning of 50. And now we like both dynamic tuning of 75.
@Manni01 with 100 real display nits prefer to have stronger hdr effect and choose a dynamic tuning of 200.
@Javs has 85 real nits and prefers also a dynamic tuning of 200.


So, the minimum/real target nits acts as a minimum so that the movie never gets brighter than intended on the uhd bluray, even for dark scenes where peak is very low.
If frame peak is 10nits, I get 50nits. And Neo-xp gets 150. And manni01 gets 100nits. And Javs gets 85nits. And Dexter Kane gets 200nits.
For all of us however, we will see this frame/scene with the exact same brightness, rendered untouched and not compressed. With the exact brightness wanted/coded in the uhd bluray.


The dynamic tuning however rules the user taste based choice:

Brightness vs hdr strength.

And it seems we all have different taste for how a good compromise is for middle scenes where peak us higher than the minimum target nits.
For bright scenes, somehow we however all agree than 2×avgHL is a good upper limit. Nobody wants to see the MEG overblown no matter what their real nits.
So here the hdr effect choice seems to win over the brightness.
And that's probably because those scene are bright to begin with, so that's ok if we offer brightness to save picture quality.

As Dexter Kane said, for someone with way more than 200 real nits, the 2×avgHL may be too low. But it seems spot on for projector and dim oled display.

:)
 

·
Registered
Joined
·
7,973 Posts
Maybe we should include the PQ value of the real display peak nits into the tuning formula?

50 nits = 0.440 PQ
100 nits = 0.508 PQ
200 nits = 0.579 PQ
400 nits = 0.653 PQ
1000 nits = 0.752 PQ
2000 nits = 0.827 PQ
4000 nits = 0.903 PQ

E.g. we could multiply the tuning parameter with "2 * PQ" value of the real display peak nits? That would result in virtually no change in the formula for projector owners. But owners of bright TVs would automatically get a higher tuning value. Maybe that way a 2000 nits TV owner could use the same tuning value as a 50 nits projector owner?
 

·
Registered
Joined
·
1,479 Posts
Discussion Starter #774 (Edited)
madmeasuredynamicclipping Tool V3.6.2

New version of our optimizer tool: V3.6.2 :)

Download:
http://projectiondream.com/download/madmeasuredynamicclipping/

New:
Focus is on the dynamic clipping to change a bit. :rolleyes:

1) Changed from scene based dynamic clipping to frame based dynamic clipping after getting positive feedback from people testing it in the LIVE algo --> Therefore more frames clipped, and more clipping strength as well for some/many frames.
-->More agressive all together but still carefull since the maximum clipping occurs only at the detected peak highlight knee. :)

2) "Clipping Strength Factor": from 0 to 1

Changed the way that the "Dynamic range recovery strength in %" is working.
- Before, you were specifying the % of pixels you wanted to clip above the detected highlight knee.
While it was based on the pixel, the issue is that is was absolutely not linear in the amount it would finally clip.

Now, this "Clipping strength factors" scales the maximum amount of clipped nits with:
Maximum clipped nits = (Original Peak - Detected Knee)

So if you have 100% you clipped everything above the knee.
If you have 50% you clipped half the nits you would have clipped with 100%.


3) "Clipping Area factor": from 0 to 1
Also, before we had hardcoded:
Do not clip if the clipped area is "too big". We had defined too big with 0.3% area.
However, it was On and OFF. With 0.29% you would fully clip, but with 0,31% you would not clip at all.

Now it is a smooth change between 0.3% and 0.5% clipped area.
Any clipping area lower or equal to 0.3% get a factor 1.
And then from 0.3 to 0.5% it smoothly change toward 0.
So that if you have 0.4% clipped surface, you would only clip 50% of the "Maximum clipped nits"
And if you have 0.5% clipped area or higher, you would not clip at all.

4) "Decompression/clipped area ratio: Factor": from 0 to 1
Before we had harcoded, if there is not enough benefit/decompression area compared to the clipped area.
Do not clip. We had put this ratio at 10:1
Again, this was a hard limit.
We have now smoothened it:
If the decompressed area is equal or more than 10 times the "clipped area", then you get of factor 1 applied on the"maximum clipped nits".
If the decompressed area is equal or less than 5 times the "clipped area", then you get of factor 0 applied on the"maximum clipped nits".
And a smooth blend in between.

5) So to sum-up the clipped peak which will be used will be:

Final frame clipped peak = Original frame Peak - (Original frame Peak - Detected Knee nits)
*(Clipping strength factor)*(Clipping Area Factor)*(Decompression factor)

6) Also, we now use the same scaling than Madshi between REC2020 peak to DCI and REC709 peak.
All the clipping, knee operation are done within rec2020 (and therefore before we just used the rec2020 values which are close anyway to the other two).

So, for more consistency, we now do:
Final frame clipped peak 2020= Final frame clipped peak

For DCI:
Final frame clipped peak DCI = (Original frame peak DCI) * (Final frame clipped peak REC2020)/(Original frame peak 2020)

For 709:
Final frame clipped peak 709 = (Original frame peak 709) * (Final frame clipped peak REC2020)/(Original frame peak 2020)



That's it.
Enjoy. :)

Anna & Flo
ps: we'll send the new code right away to you @madshi ;)
 

·
Registered
Joined
·
1,133 Posts
So did I just perform a monumental waste of time? I went ahead and downgraded from Windows 10 to Windows 7.

edit: Yea I'm dumb and should've gone down to Win8.1. I'm going to reinstall Win10 and leave it at that. 10 is a bit more stable for gaming (steam) anyway
 

·
Premium Member
Joined
·
9,901 Posts
@Neo-XP

Please could you post your current settings for scene/chapter change with the tool?

I'd like to try your approach here as well after becoming a convert with the live algo in the other thread :)

I need to recalc my test files for my new peak nits (120nits), so it would be a good opportunity to try.
 

·
Registered
Joined
·
821 Posts
I need to recalc my test files for my new peak nits (120nits), so it would be a good opportunity to try.
Off: how is your new projector? :) (compared to the previous one)
 

·
Premium Member
Joined
·
9,901 Posts
Off: how is your new projector? :) (compared to the previous one)
I'll soon post recommended settings, tips and lots of measurements in the new calibration thread for the new models (google my sig and subscribe to the thread, I'll post in that thread whenever I add content in the reserved first posts). I'll also probably post some first impressions in the owners' thread there at some point, but I need to set-up and calibrate the beast (even if minimally) before I can watch content and talk about PQ. I'm very happy with it for now though, if you want the tldnr :)

Back to topic :)
 

·
Premium Member
Joined
·
1,760 Posts
New version of our optimizer tool: V3.6.2 :)

Download:
http://projectiondream.com/download/madmeasuredynamicclipping/

New:
Focus is on the dynamic clipping to change a bit. :rolleyes:

1) Changed from scene based dynamic clipping to frame based dynamic clipping after getting positive feedback from people testing it in the LIVE algo --> Therefore more frames clipped, and more clipping strength as well for some/many frames.
-->More agressive all together but still carefull since the maximum clipping occurs only at the detected peak highlight knee. :)

2) "Clipping Strength Factor": from 0 to 1

Changed the way that the "Dynamic range recovery strength in %" is working.
- Before, you were specifying the % of pixels you wanted to clip above the detected highlight knee.
While it was based on the pixel, the issue is that is was absolutely not linear in the amount it would finally clip.

Now, this "Clipping strength factors" scales the maximum amount of clipped nits with:
Maximum clipped nits = (Original Peak - Detected Knee)

So if you have 100% you clipped everything above the knee.
If you have 50% you clipped half the nits you would have clipped with 100%.


3) "Clipping Area factor": from 0 to 1
Also, before we had hardcoded:
Do not clip if the clipped area is "too big". We had defined too big with 0.3% area.
However, it was On and OFF. With 0.29% you would fully clip, but with 0,31% you would not clip at all.

Now it is a smooth change between 0.3% and 0.5% clipped area.
Any clipping area lower or equal to 0.3% get a factor 1.
And then from 0.3 to 0.5% it smoothly change toward 0.
So that if you have 0.4% clipped surface, you would only clip 50% of the "Maximum clipped nits"
And if you have 0.5% clipped area or higher, you would not clip at all.

4) "Decompression/clipped area ratio: Factor": from 0 to 1
Before we had harcoded, if there is not enough benefit/decompression area compared to the clipped area.
Do not clip. We had put this ratio at 10:1
Again, this was a hard limit.
We have now smoothened it:
If the decompressed area is equal or more than 10 times the "clipped area", then you get of factor 1 applied on the"maximum clipped nits".
If the decompressed area is equal or less than 5 times the "clipped area", then you get of factor 0 applied on the"maximum clipped nits".
And a smooth blend in between.

5) So to sum-up the clipped peak which will be used will be:

Final frame clipped peak = Original frame Peak - (Original frame Peak - Detected Knee nits)
*(Clipping strength factor)*(Clipping Area Factor)*(Decompression factor)

6) Also, we now use the same scaling than Madshi between REC2020 peak to DCI and REC709 peak.
All the clipping, knee operation are done within rec2020 (and therefore before we just used the rec2020 values which are close anyway to the other two).

So, for more consistency, we now do:
Final frame clipped peak 2020= Final frame clipped peak

For DCI:
Final frame clipped peak DCI = (Original frame peak DCI) * (Final frame clipped peak REC2020)/(Original frame peak 2020)

For 709:
Final frame clipped peak 709 = (Original frame peak 709) * (Final frame clipped peak REC2020)/(Original frame peak 2020)



That's it.
Enjoy. :)

Anna & Flo
ps: we'll send the new code right away to you @madshi ;)

With this new version, using the default settings (other than 105 real display nits and 125 dynamic tuning) I am seeing some huge fluctuations in brightness.

For example, Oceans Eight, chapter 6 (Vogue offices) there is a conversation that takes place between two women. Every time it changes to the women who is conducting the interview, the brightness changes. One time it is dark, the next it is bright, then dark, then bright etc.
 
761 - 780 of 1220 Posts
Top