AVS Forum banner

6681 - 6700 of 9410 Posts

·
Premium Member
Joined
·
727 Posts
I don't really know what's happening here, but I love it:

HSTM Off / HSTM On (my settings) / HSTM On (my settings + start at:100)


LG Colors of Journey HDR UHD 4K Demo - Frame 4511

Holy cow that reminds me of what my old Darbee used to do.
 

·
Registered
Joined
·
1,208 Posts
Thank you for this feedback. I've actually found myself really enjoying 15-30-15 the past couple days in case anyone wants to try it. It gives a nice balance between depth and pop (perceived).
I tested your settings, but I get a tiny bit more depth with 20-40-20:

15-30-15 / 20-40-20


The Meg - Frame 96903

I agree that the difference is very small, but still visible ;)
 

·
Registered
Joined
·
1,479 Posts
Discussion Starter #6,683
Did someone tried to use more than 100% strength with the default HSTM settings yet?

Madshi said the current 100% strength was really jut a random first strength he had chosen for the scale.

200%, 300%,... 500% anything could still be a good "theoretical " value. Would be nice to compare (and I can't right now :D but I am still curious! :rolleyes: )
 

·
Registered
Joined
·
410 Posts
I'm still not 100% sure about my preferred settings yet...

Just looking at the start setting...

With Atomic Blonde increasing the start values gradually improves the image until around 175 (which might be a fraction better than 150). At 200 it clips the detail in the hair. The measured frame peak is around 315 iirc

350 static target nits (htsm off)






i.e. 0 vs 50 vs 100 vs 150 vs 200




I do like the higher start setting in other scenes...

Lucy loses some APL but emphasises highlights


0 vs 150





0 vs 100



The area around her cheek/right eye is better defined with 100 imo










but there are some scenes where 0 is better than 100 (& 150) imo

0 vs 100 vs 150




0 seems to have better depth and the balance of the light is better imo

Again the same thing with the image below imo

0 vs 150







I'm also looking at 100 strength and 20-40-20 vs 50 strength and 10-20-10 and again some scenes can have more contrast than others...

Neo's Meg scene is one in favour of 100 strength, 100 start and 20-40-20, but in this example I think 50 strength, 100 start and 10-20-10 has more depth





Also, I think the mountain top might have a bit more depth/ stand out more (?)



:confused:

Will still need to do more testing...
 

·
Registered
Joined
·
1,208 Posts
@Fer15

This scene of The Meg look awful with my settings when a "start at nits" value is applied:

HSTM Off


The Meg - Frame 28753

HSTM On 100-20-40-20 / HSTM On 50-10-20-10



50-10-20-10 has clearly more depth than 100-20-40-20, but the colors are wrong IMO.

However look at what happens when a start at nits value of 100 is applied:

HSTM On 100-20-40-20 + start at nits 100 / HSTM On 50-10-20-10 + start at nits 100



100-20-40-20 looks awful, not 50-10-20-10.

I will try 50-10-20-10 on other scenes ;)

Thanks Fer15!

Edit: The maximum "start at nits" value would be 10 with my settings:

HSTM On 100-20-40-20 + start at nits 10



Edit2:

HSTM Off / HSTM On 100-20-40-20 + start at nits 0 / HSTM On 50-10-20-10 + start at nits 100



Yep, 50-10-20-10 + start at nits 100 looks better to me :cool:
 

·
Registered
Joined
·
7,948 Posts
b98, if I use the attached settings (ouput HDR format) and disable "dynamic clipping" then sometimes I get lower targets (brighter image): does it make sense?
I thought "dynamic clipping" "always" result in lower targets.
Used on its own it should be the case. When using HSTM (Histrogram shaped tone mapping), I would expect the same, but I'm not 100% sure, the exact behaviour can be a tad unpredictable at times. FWIW, when using HSTM, a lower target doesn't have to result in a brighter image. When using HSTM, you can't really tell much by looking at the target nits.

The skin color is 1000x more natural with the "start at" option, IMO.

The "start at" value will surely not be the same for every content. Maybe based on FALL or something else... (because measured peak is not reliable for this and can change very fast).

But this option alone could be the topic of another test, with its own set of parameters...

Edit: Another test:

Edit2: a value of 100 seems to be already beneficial in most cases?
Aren't we back to judging saturation here, though? Which I asked you to ignore (for now)?

I don't really know what's happening here, but I love it:

HSTM Off / HSTM On (my settings) / HSTM On (my settings + start at:100)

LG Colors of Journey HDR UHD 4K Demo - Frame 4511
I guess the detail in the stone wall looks better, but it's hard to say how it's supposed to look.

I'm still not 100% sure about my preferred settings yet...

Just looking at the start setting...

With Atomic Blonde increasing the start values gradually improves the image until around 175 (which might be a fraction better than 150). At 200 it clips the detail in the hair. The measured frame peak is around 315 iirc
I think you're mostly just using different target nits values here and judge the results.

50-10-20-10 has clearly more depth than 100-20-40-20, but the colors are wrong IMO.
Why are we discussing saturation? :(
 

·
Registered
Joined
·
7,948 Posts
Here's the next test build:

http://madshi.net/madVRhdrMeasure99.zip

1) I've modified the desaturation option. You now now choose to apply no desaturation at all, and I'd kindly ask you to use this option for testing (but not for actual playback, I think). This way hopefully you'll be able to judge different HSTM settings better without being influenced by different saturation levels. Well, at least I hope so... :p

2) Because I thought we didn't have enough options yet, I've added a couple more... :eek::(

3) The "start at nits" option is now split into "don't compress below" and "calculate FALL above", which are both in nits. Would be interesting to check which of these is the one which you think helps quality. I anticipate that it might be the "calculate FALL above" option, which would mean that it basically only influences madVR's decisions on how to bright make the final image (and nothing else).

4) Shadow boost has 2 even stronger settings now (Neo-XP asked for it for testing purposes).

5) There are two more options which let you choose two different areas of behaviour for HSTM. These two options let you go back (to some extent) to the way build 96 worked, just to double check if that looks better or worse than 97. And also to double check if the "blow out" problems with the original 96 build still exists or maybe it was fixed by some of the other improvements I made. Soulnight asked for these options.

Let me know what you think!! :)
 

·
Registered
Joined
·
502 Posts
Strange! Any ideas why? In which way does it look better?
LG HDR monitor looks better than JVC projector for HDR content, but...(Projector is gaining with v96).

I wasn’t very clear (deleted my post). My point was to say with v96 and outputting SDR2020 “tone mapped“, JVC projector looks much closer to my LG 400 nit monitor which I always run in HDR mode, when watching the same movies. HSTM=100% and all other new settings at default, seems to preserve the HDR picture elements better when outputting SDR2020. A step up in perceived picture quality. You should watch the video in motion, not just single frame compares, to see the full effect. I would not want to loose it. Maybe others can confirm? Fingers crossed HSTM makes it to final-cut. Thanks.:)
 

·
Premium Member
Joined
·
16,949 Posts
I would like to comment on the color. Just gave 99b a test drive. My projector ( RGB Laser ) is capable of beyond P3. With desaturation chosen color is still way under saturated, 709 at best, not seeing any difference between desat and 1 or 2. I believe the JVC projectors can do P3, is no one else running P3?
Color aside other aspects of build 99b excel. Thank you madshi :)
 

·
Registered
Joined
·
1,208 Posts
I can't get this scene "right" with build 96 parameters:

HSTM On build 96 parameters (default settings) / HSTM On build 97 parameters (default settings)


The Meg - Frame 9384

In grayscale :p:

HSTM On build 96 parameters (default settings) / HSTM On build 97 parameters (default settings)



Edit: Same as before, the only way is to fix it is to change the clipping point to 0:

HSTM On build 96 parameters (default settings + clipping point: 0)

 

·
Registered
Joined
·
7,948 Posts
Fingers crossed HSTM makes it to final-cut. Thanks.:)
It seems very likely at this point.

I would like to comment on the color. Just gave 99b a test drive. My projector ( RGB Laser ) is capable of beyond P3. With desaturation chosen color is still way under saturated, 709 at best, not seeing any difference between desat and 1 or 2. I believe the JVC projectors can do P3, is no one else running P3?
Soulnight and Manni are running P3, if I'm not mistaken.

Hmmmm... Two questions to "debug" this saturation issue you're seeing:

1) Is this a new problem with the most recent test builds? Or did you always have this problem with madVR that colors are way under saturated? If it's a new problem, can you say which is the last madVR build which doesn't have this problem?

2) Are you running a 3DLUT, or are you using "this display is already calibrated"? In the latter case, do you see any difference in saturation when you switch the "this display is calibrated to" gamut to different values?

Hey guys, had a quick look at HSTM in the new build - these settings seem ok:
https://ibb.co/2Mrc2Gs
I think these are not far from what Neo-XP and Fer15 are using, although some things are still under discussion, e.g. the max sky width, sky strength and the HSTM power/clipping point/strength.

I can't get this scene "right" with build 96 parameters:

HSTM On build 96 parameters (default settings) / HSTM On build 97 parameters (default settings)

Same in grayscale :p:

HSTM On build 96 parameters (default settings) / HSTM On build 97 parameters (default settings)
Haha! I had actually considered to add a grayscale option to the desaturation combobox... :D
 

·
Registered
Joined
·
1,208 Posts
3) The "start at nits" option is now split into "don't compress below" and "calculate FALL above", which are both in nits. Would be interesting to check which of these is the one which you think helps quality. I anticipate that it might be the "calculate FALL above" option, which would mean that it basically only influences madVR's decisions on how to bright make the final image (and nothing else).
don't compress below 0 / don't compress below 100


Pacific Rim - Frame 173097

Is the "don't compress below" option working as intended? Shouldn't the blacks be less compressed with 100? I used no shadow boost here.
 

·
Registered
Joined
·
7,948 Posts
don't compress below 0 / don't compress below 100

Pacific Rim - Frame 173097

Is the "don't compress below" option working as intended? Shouldn't the blacks be less compressed with 100? I used no shadow boost here.
Good question. I'm not fully sure how all the options interact with each other. For example, if you tell HSTM to try to keep the FALL the same, then not compressing the area from 0-100 may mean that HSTM has to compress *everything* stronger to keep the FALL the same, which may help with 20-100 nits, but could actually harm 0-20 nits. But I've no idea. There are so many different options which can interact with each other in complicated ways, so I'm not sure.

FWIW, "don't compress" means that HSTM shouldn't apply *additional* compression on top of the linear compression which happens with a higher target nits. So if telling HSTM to not compress below 100 nits results in a higher target nits to be used, this can counter-act the "don't compress" option in that way that HSTM doesn't apply additional compression below 100 nits, but to the higher target nits, the shadow area might still be compressed stronger overall due to the higher linear compression cause by the higher target nits.

Hope that makes sense?
 

·
Registered
Joined
·
1,208 Posts
Hope that makes sense?
Yes, I think I will let this setting alone for now :)

However, the other one, "calculate FALL above", seems very beneficial on scenes like this with very bright highlights on faces:

calculate FALL above 0 / calculate FALL above 100


The Meg - Frame 116231

Question: what's the difference between "curve" and "limit curve"? They are my favorites to keep good targets, but I see very little change between the two.
 

·
Registered
Joined
·
31 Posts
Hey guys, had a quick look at HSTM in the new build - these settings seem ok:
https://ibb.co/2Mrc2Gs



Thanks to all involved in development and testing.
Just saying : using the last build and the above settings (except real display peak nits set to 130 which is what I get on my Sony HW40ES), I experienced unwanted brightness changes watching The Lion King (2019), as if scene detection was messed up.


I used my wife's bathroom break to quickly revert back to whatever the last madvr test version was before madshi went on a hiatus to work on the Envy and it was fine... :D


I don't have specific timestamps (sorry...) but I know you guys have special clips to test sudden brightness changes so maybe you should run a few tests to be sure everything works as intended ;)
 

·
Registered
Joined
·
13 Posts
Hello Madshi,
am unfortunately only a lay tester at madVR. But maybe my opinion is valuable?

Have a JVC X7900 and come a year ago from the Panasonic Blu-ray Player to madVR. I have always missed the saturated colors. I think I'm not alone.
Now it is very good with these settings (99b).:)
JVC DCI calibrated Low Lamp Gamma 2.2



But more important as the colors is the depth of the picture. The comes in variant V96 very special to well. It's like Soulnight says like a veil that goes away. Or like putting on glasses. And in contrast to the real V96, the skin colors fit again. So please, please follow this approach.
Photos are with smartphone of 3.10m screen 55Nits.

The following things I have noticed negatively.
Here and there flickering. Is known and will probably be fixed.
Highlights cliffs partially and look a bit washed out. But it can also be that I do not get better with 99 b set or made a mistake there?

I look forward to the special V100.:)
 

Attachments

·
Registered
Joined
·
1,479 Posts
Discussion Starter #6,699 (Edited)
@madshi thanks for the new build with all those nice options.

As you know, I could not test over the weekend but I managed to test a bit this evening with Anna.

5 profiles:
1) Htsm off: desaturation 1+2
2) Htsm off: desaturation 1
3) Htsm off: no desaturation
4) Htsm 100% strength: 96 + no desaturation
5) Htsm 100% strength: 97 + no desaturation

We tested a few scenes of the beginning "guardian of the galaxy 2"

First, I saw no scenes where no desaturation looked bad. In fact, in many scene it was a welcome natural uptick on color compared to desaturate 1, and a huge one to desaturate 1+2 without any apparent highlights details.
So at least on what I tested, no desaturation was clearly my favorite.

Then I proceeded to compare profile 3) 4) 5) all with "no desaturation "

Both 4) and 5) were always a nice improvement in picture clarity compared to 3. I love hstm. :)

Since 4) with 96 is often brighter, it was a bit difficult to compare....
But to be fair, I found today than often 5) with 97 looked better than 4) with 96.

Having said that, the opposite also happened sometimes with 96) looking better than 97) and that without being any brighter (so not the reason this time). In fact sometimes 97) looks darker than htsm off, like the fall correction would not be working properly. How do you proceed to ensure brightness match exactly?

And I had still the feeling sometimes that 97 was robbing some saturation compared to htsm off.

I will provide some screenshots hopefully by tomorrow to document this so that we can try to take the best from 96 and 97.

Thanks for the hard work!
Florian :)
 

·
Registered
Joined
·
7,948 Posts
However, the other one, "calculate FALL above", seems very beneficial on scenes like this with very bright highlights on faces:

calculate FALL above 0 / calculate FALL above 100

The Meg - Frame 116231
FWIW, this isn't really a key part of HSTM. Basically this "calculate FALL above" option modifies the original FALL algo a bit to pick a different target brightness/peak to render at. If you use a nits number above 0, you're telling madVR to ignore a part of the histogram in its decision making. I'm not fully sure what the practical effect will be. I can imagine it to be somewhat unpredictable because it might be sort of "random" whether the part of the histogram you're igoring is higher or lower than the average. So it could make the image either darker or brighter, and I'm not sure if the changes that are done this way are scientifically sound. But I guess proof is in the pudding, so if after careful testing you think you get better decisions this way, that's fine with me.

Question: what's the difference between "curve" and "limit curve"? They are my favorites to keep good targets, but I see very little change between the two.
"curve" doesn't have any limits. "limit curve" means that madVR doesn't allow the HSTM target nits to be below the real display nits. So this option will only produce different results in situations where "curve" would produce a target nits level that is below your real display nits.

I experienced unwanted brightness changes watching The Lion King (2019)
That is as expected. Please stick to test build 95 for real movie watching use. The newer test builds are only for testing purposes.

am unfortunately only a lay tester at madVR. But maybe my opinion is valuable?

Have a JVC X7900 and come a year ago from the Panasonic Blu-ray Player to madVR. I have always missed the saturated colors. I think I'm not alone.
Now it is very good with these settings (99b).:)
JVC DCI calibrated Low Lamp Gamma 2.2
I've said it like a million time in the last couple of pages: We will revisit saturation later. I'm not interested in discussing it now. I'm only interested in discussing HSTM for now. Which should have nothing to do with saturation. Please use "don't desaturate" when testing HSTM, this way HSTM should not influence saturation.

I *am* interested in HSTM, and I'm interested in your opinion there, if you really test various options yourself and can explain which options you prefer for which reason.

But more important as the colors is the depth of the picture. The comes in variant V96 very special to well. It's like Soulnight says like a veil that goes away. Or like putting on glasses. And in contrast to the real V96, the skin colors fit again.
So which setting combinations for the 2 new drop-down-boxes on the bottom right part of the HDR settings page did you compare? I suppose you compared 96/96 to 97/97? Did you try anything else? E.g. 96/97 or 97/96? And which settings combinations did you prefer for which reason?

Highlights cliffs partially and look a bit washed out.
Did this happen with all settings combinations you tried or only with some?

First, I saw no scenes where no desaturation looked bad. In fact, in many scene it was a welcome natural uptick on color compared to desaturate 1, and a huge one to desaturate 1+2 without any apparent highlights details.
So at least on what I tested, no desaturation was clearly my favorite.
Which is all nice and fine, but as you know, I've zero interest in discussing saturation at this point in time. There will be a time to discuss saturation in the near future, but not now. For now I will ignore any and all comments related to saturation, I'm sorry.

Then I proceeded to compare profile 3) 4) 5) all with "no desaturation "
Yes, please... :):)

Both 4) and 5) were always a nice improvement in picture clarity compared to 3. I love hstm. :)

Since 4) with 96 is often brighter, it was a bit difficult to compare....
But to be fair, I found today than often 5) with 97 looked better than 4) with 96.
I suppose you compared both drop-down-boxes set to 96 with both set to 97, correct? Might be interesting to also try mixes of both, or try the other options which are neither 96 nor 97.

FWIW, the first of the 2 new drop-down-boxes basically only changes the brightness madVR renders the image at. So you can think of this as different mods to the FALL algorithm. While the 2nd drop-down-box changes the way the HSTM algorithm tries to achieve the goal defined by the first drop-down-box. So the 2nd drop-down-box is probably more important (?).

Having said that, the opposite also happened sometimes with 96) looking better than 97) and that without being any brighter (so not the reason this time).
That could be quite interesting to analyze in more detail.

In fact sometimes 97) looks darker than htsm off, like the fall correction would not be working properly. How do you proceed to ensure brightness match exactly?
If my math is correct, I'm trying to achieve the same FALL value *after* tone-mapping, with HSTM turned on vs off. Of course this only applies if you set the first of the 2 new drop-down-boxes to "keep FALL".

I will provide some screenshots hopefully by tomorrow to document this so that we can try to take the best from 96 and 97.
Looking forward to that. Please also consider testing the 2 new drop down boxes set to a mix of 96/97 or 97/96. E.g. maybe one of the drop-down-boxes has a very clear winner, and maybe the other one doesn't.
 
6681 - 6700 of 9410 Posts
Top