AVS Forum banner
201 - 220 of 1238 Posts

· Registered
Joined
·
1,482 Posts
Discussion Starter · #201 · (Edited)
But then, of course, it means that it's not easy to find a target that works for the whole chapter. It has to be a compromise.

Anyway, having abrupt changes of target just because there is a bright light in a shot or a flash or an explosion isn't what I was after. Maybe along %FALL we need a minimum duration (in frames) to cancel the change if FALL gets back below the threshold within that amount of time.

For example, if we had a WINDOW value of say 120 frames (5 seconds, adjustable of course), we could avoid the changes everytime there is a brief flash or explosion, but when the changes are for longer, we could change the target instantaneously when the "chapter" starts. This would allow to have a lower %FALL value, so we do have changes when there is a change in each chapter, but we don't have chapters of less than 120 frames.

This might mean that you need two passes: one to define the chapters (and no reset if the change above %FALL is not longer than WINDOW), another to calculate the best target for each chapter (so that we find the target that gives the best balance between highlights and shadow detail), once you have identified them.

Thoughts?
Hi :)

Well, I definitely don't like jump either or abrupt change at the wrong time either. That's why I went with a large (10s) rolling avg in the first place.

Then @Neo-XP wanted to see if resetting the target nits at each scene change would improve the global experience and if not, what could we do to improve it.

So, we tried to implement that.

Since flo scene was to jumpy to our taste, we tried to improve upon that.
We thought, for sure, nobody will be happy resetting the target nits at each "original madshi scene definition", so no reason to release a version with only that algo in addition to "Flo".


So to improve, we need to filter the scenes so we would keep only scenes/chapters in between a large change is occurring.
That's what we did in the current chapter definition.

So my thoughts are that your current proposals make a lot of sense to get what we desire as a "target nits" chapter and further filter down and combine the original scenes. :)

I fully agree with you that we need to introduce even more rules to filter out even more scenes.

I also thought that a minimum chapter time duration was a good idea so we'll try that next. :)

And then we would need to compare the "avgFALL over a complete chapter" to the next. If this is above a selected threshold, then the 2 chapters would stay separated. If not they would merge into 1. :)
Or something like that.

@Dexter Kane
I also thought yesterday evening that looking at something called the avgLL (avg low light 0-100nits) to look for Chapter changes might be a good option.
We will probably try that as well. ;)
 

· Registered
Joined
·
1,440 Posts
IMO, the current algo is relaying too much on highlights to define the target, therefore the target is often too high and changing too much from scene to scene.

For instance, if highlights represent 10% of an image, then they should only count somehow only for 10% on the target formula.

When two people are discussing together in the same room, the average brightness should remain approximately the same, even if the camera is switching between them.
 

· Premium Member
Joined
·
11,304 Posts
I agree with @Dexter Kane and @Neo-XP. :)

As you know, I always wanted 0-100nits stable. If not over the whole film, at least during a chapter. Maybe looking at 0-100nits for chapter detection (but not necessarily for target calculation) or giving it a much higher weight would be the way to go? Or a combination with the window idea?

Thoughts?
 

· Registered
Joined
·
1,482 Posts
Discussion Starter · #204 · (Edited)
So the new strategy algo to select proper chapter should look like that:


Step1: detect chapter with %FALL change (user choice)

Step2: combined chapter to have time duration > 5s (user choice)

Step3: based on latest chapter definition, combine chapter if change from chapter (k) to chapter (k+1) for avgFALL less than x%
Edit: step 3 could also use avgLL (0-100nits) instead of avgFALL.

This should eliminate a good number of cases where we got unwanted reset of the target nits with "flo chapter".

@Neo-XP this should also solve the issue you have with scenes and people talking in the same room ;-)

Afterwards if people want to have constant target nits over a chapter or a moving rolling avg can be decided with the rolling avg time. Huge number= constant , small number (240) moving.


Goal is to have the same smoothness as "flo" algo (which never jumped, especially not for people in the same room) but detecting REAL chapter cut with big brightness difference so that the rolling avg can be resetted instead of displaying fast target nits change.

So it should look like "flo" in better. Flo scene does not makes sense to me.
 

· Registered
Joined
·
1,440 Posts
Another scene looking way too dark because of the highlights in the background:

Dynamic Target Nits Tuning 50 / 75 / 100


La La Land - Frame 146572

50 is ok to me here, but then other scenes look too bright :confused:

This happens almost every time there are some lights in the frame, a character in front of a big window of screen, the sky in a part of the image, etc.

Apart from that, I don't have any issue with the current "Flo Chapter" algo.
 

· Registered
Joined
·
1,482 Posts
Discussion Starter · #207 ·
Another scene looking way too dark because of the highlights in the background:

Dynamic Target Nits Tuning 50 / 75 / 100


La La Land - Frame 146572

50 is ok to me here, but then other scenes look too bright :confused:

This happens almost every time there are some lights in the frame, a character in front of a big window of screen, the sky in a part of the image, etc.

Apart from that, I don't have any issue with the current "Flo Chapter" algo.
How does that shot looks with standard "flo" without scene or chapter and rolling avg 240 or higher?
 

· Registered
Joined
·
1,440 Posts
How does that shot looks with standard "flo" without scene or chapter and rolling avg 240 or higher?
Exactly the same:

Dynamic Target Nits Tuning 50 / 75 / 100



It is not an issue introduced by the "Flo Chapter" algo, which is working very well for me now with 100% FALL change in % :nerd:
I guess the problem is that the algo is based on the avg nits of highlights, hence not talking into consideration the % of pixels of 100 nits and below.
 

· Premium Member
Joined
·
1,973 Posts
Another scene looking way too dark because of the highlights in the background:

Dynamic Target Nits Tuning 50 / 75 / 100


La La Land - Frame 146572

50 is ok to me here, but then other scenes look too bright :confused:

This happens almost every time there are some lights in the frame, a character in front of a big window of screen, the sky in a part of the image, etc.

Apart from that, I don't have any issue with the current "Flo Chapter" algo.
This is something that I've been struggling with (note that I haven't tried Flo's latest version). If the background is bright, foreground objects that aren't overly bright (e.g. people) can appear dark - when in reality they are the subject of the scene, not the background. It's similar to the silhouette effect of having the sun facing you - your eyes can't properly see details of foreground objects.

In trying to reduce this effect, it's difficult to find a target nits value that doesn't make the entire image too dull. I am currently using 87 but find I like slightly varied values )between 85 and 90) depending on the movie.
 

· Registered
Joined
·
2,039 Posts
Umm, 32-bit LAV Splitter Source + LAV Video Decoder?
What about 64-bit? I only installed those.


Cheers
Tom
You have to install 32-bit LAV Filters. The standard installer will install both 32-bit and 64-bit versions, so you might have them installed already.
 

· Registered
Joined
·
2,039 Posts
Thanks.... in case of support how can you interact that post ? would not it be more useful for those who need support to open a thread? unfortunately in this thread I asked for some information , without receiving satisfaction or support.sometimes I wonder if it's just for users super prepared this thread . If some things seem trivial to most people for me or someone else they are not.

I would like to close with a thanks to Soulnight for the fantastic tool that made us available. I will try to understand its functioning elsewhere. For me, writing in English is very difficult. Without controversy.


Warner306, your guides that you publish on the Kodi forum , they are true notions of how a user medium and not only he can really learn.
You could ask your question here. Maybe it was missed or misunderstood. I'll move a dozen guides relating to madVR to AVS Forum at some point, but not in the next couple of weeks.
 

· Registered
Joined
·
1,440 Posts
Could you provide me examples for those too bright frames? ;)
Sure:



The first 3 are from 00:26:07-00:26:20 runtime, the 4th one is frame 142666.

Edit:

Also this one for instance from The Meg:


Frame 61280

Basically, most frames without bright highlights that make the target go much higher.

With Dynamic Target Nits Tuning at 100, those frames are ok, but then frames with some % of bright highlights are too dark :devil:
 

· Registered
Joined
·
1,482 Posts
Discussion Starter · #215 ·
Sure:



The first 3 are from 00:26:07-00:26:20 runtime, the 4th one is frame 142666.
I don't see any issue there. :)
Target is close to the peak nits even.

With 100, target nits will not be much higher since the algo pay attention not to have a target nits above the peak nits.

Also you should use your measured display peak as the minimum value.
You said you had 150nits. So please use that. Otherwise if you use a minimum target lower than your real display nits, you are trying to visualize the content brighter than it was intended for.
 

· Registered
Joined
·
116 Posts
You could ask your question here. Maybe it was missed or misunderstood. I'll move a dozen guides relating to madVR to AVS Forum at some point, but not in the next couple of weeks.

Thanks you for answering , then , I try to ask the question directly to you and I take inspiration extrapolating a section of your guides :

Configuring madVR's Tone Mapping

target peak nits: [200] Peak nits is the target display brightness for tone mapping in PQ nits. The value is not meant to correlate to the actual brightness of the display when converted to SDR gamma. HDR content requires you adapt the source to your display, so choose any value that looks best to you. To create dynamic range, increase the nits target above the actual peak nits of the display to add contrast. This will best preserve the HDR presentation and give the image greater detail and depth.


I highlighted the steps!

the question is

if my TV panel to actual brightness of the display measured about 700 nits , in your opinion , what value should I enter in the GUI of madVR "target peak nits: " ?
 

· Registered
Joined
·
1,440 Posts
I don't see any issue there. :)
Target is close to the peak nits even.

With 100, target nits will not be much higher since the algo pay attention not to have a target nits above the peak nits.

Also you should use your measured display peak as the minimum value.
You said you had 150nits. So please use that. Otherwise if you use a minimum target lower than your real display nits, you are trying to visualize the content brighter than it was intended for.
Those frames are ok with 100, too bright with 50. Of course I can set 150 for the minimum target, but I have this issue with all titles and not limited to the 100-150 target.

For instance, the frame from The Meg with 100 Dynamic Target Tuning instead of 50:



It looks much better.
 

· Registered
Joined
·
1,482 Posts
Discussion Starter · #219 ·
Those frames are ok with 100, too bright with 50. Of course I can set 150 for the minimum target, but I have this issue with all titles and not limited to the 100-150 target.

For instance, the frame from The Meg with 100 Dynamic Target Tuning instead of 50:



It looks much better.
:)

I am really for improving the dynamic algo as much as possible.

But some things are also just related to user taste. :)

The examples you showed in lalaland will definitely look much better on your bright 150nits screen if you use your 150 real nits as minimal nits.
So this will solve some of the
frames you dislike with algo 1.
(And again, I they are already very close to the frame peak nits each time and look perfectly fine to me. Again algo 2 will bring very little for those. Only your minimum target will reall help).

For the meg frame, again it's discutable.
It seems that algo 1: avgHL + avgFALL does not give you high enough a target nits for your taste on this frame.

But algo 2: 2 times avgHL does.

At the same time, you say that the algo focuses too much on the highlights and should look at the whole picture.
Well doing that here would get your an even lower target nits.

You can see with your own eyes that the highlights avg is not that high in that frame.

So here you would like the peak highlights to be more relevant. And in others frame like Lucy, you ask exactly the opposite
;)

I was planning to bring a beta fix for your Lucy example taking more in consideration the rest of the picture.
But if I use it also for this frame from Meg, algo 2 (100) will also get lower.

Personally I think the core algo is already quite balanced.

To do much better, we would need to plug neo-xp taste in a AI neural network :)

Joke aside, let see what we can do. Chapter update may also help in that regard.
 

· Registered
Joined
·
1,440 Posts
Well most of the time algo 2 (100) looks just fine to me, I think it just needs to be adapted a bit to be really usable and not look too dark on some conditions only (not dependent to user taste).

I don't know if that is possible without making anything worse though :D
And if it is possible, we must find a way ;)

Same as the previous example, 100 is the only valid choice here IMO:

Dynamic Target Nits Tuning 50 / 75 / 100


The Meg - Frame 61547
 
201 - 220 of 1238 Posts
Top