Jump to content


Photo

HDR tone-mapping with mpv


Best Answer Doofus , 19 March 2019 - 09:45 PM

Here's a starter pack for those who haven't done this, before. This assumes that your desktop coloring looks like a standard desktop. This is important!

1. Download the Adobe attachment, and unzip it to; C:\Windows\System32\spool\drivers\color

2. Download the mpv config attachment, and unzip it to; %AppData%\Roaming

3. Go play a movie :)

Any questions, ask them, here :)

 

(more information and an updated configuration, can be found here)

Go to the full post


  • Please log in to reply
295 replies to this topic

#261 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 06 September 2019 - 11:04 PM

This looks interesting. Once it gets rolled into the git build, I'll see what difference it yields.

 

https://github.com/m...e6177a57f81785a

 

5d731e1296f56_targetpeak.jpg



#262 lorac OFFLINE  

lorac

    Advanced Member

  • Members
  • 409 posts
  • Local time: 12:13 AM
  • LocationVancouver, Canada

Posted 07 September 2019 - 12:04 AM

Reads like a different language. If you find it interesting though I'm sure it's good.

Sent from my ONEPLUS A6003 using Tapatalk

#263 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 07 September 2019 - 12:21 AM

It's basically saying that it's messing with how the shaders work, so its result is the wrong value for peak luminance.


Edited by Doofus, 07 September 2019 - 01:00 AM.


#264 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 09 September 2019 - 01:13 PM

My colorimeter was delivered to someone in another state :/

#265 markyp OFFLINE  

markyp

    Advanced Member

  • Members
  • 238 posts
  • Local time: 08:13 AM

Posted 09 September 2019 - 01:15 PM

Oh dear!

#266 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 09 September 2019 - 01:20 PM

I know, right? I'm waiting to hear from the shipper. It's through Amazon, so it should get rectified....at some point.

#267 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 14 September 2019 - 06:36 PM

I'm messing with my colorimeter and using DisplayCal. Bit of a learning curve, when calibrating for WCG. I bailed on the software that comes with it. Clearly that is designed for regular computers. Difficult to use on my system. DisplayCal is easier, and more complicated at the same time. I'm still getting to grips with it. I'm on my 4th run through...

Edited by Doofus, 14 September 2019 - 06:36 PM.


#268 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 14 September 2019 - 08:31 PM

It doesn't look like I'm going to get what I want. At best, I think I may only get slight improvement over what I already have. That is to say only slightly more accurate lighting. Not even noticable if you don't know what to look for.

#269 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 15 September 2019 - 03:43 PM

So, apparently when calibrating a wide gamut display, it's better to use a spectrometer instead of a colorimeter. Using a colorimeter, it requires color correction. If there is one available for your display, then you'll be fine. But if like me, you don't, crap! The i1 apparently can compensate, but I have yet to figure out how :/

#270 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 15 September 2019 - 05:00 PM

I'm gaining ground. Huzzah!

#271 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 15 September 2019 - 08:03 PM

In case anyone is interested, this is what I've been wrestling with.

 

https://www.argyllcm...utColmters.html

 

The i1 does appear to be able to correct for this, if you let it make it's own measurements. I decided to use DisplayCal. It's free, and appears to be pretty good, though I don't have much to compare it to. After delving into the advanced settings, I did eventually produce a decent ICC profile, and I am now using it with mpv. The results are pretty good, but I think it's mostly due to the calibration and not the ICC (which is built on said calibration). Color accuracy is good, but I could have used much larger color patches to gain greater accuracy. I just didn't have the patience to wait more than an hour. Considering the i1 is apparently the fastest colorimeter, I imaging that other meters must take many more hours to complete the same task. 

 

I made several profiles, and tested each one. I finally figured out how to make a wide gamut profile, and made two profiles with that. One with a 3dLUT and one without, keeping all the other settings the same. There was a notable difference in color accuracy between them. The one with a 3dLUT being the better of the two. This is what I am using with mpv. It also appears as those with 3dLUT, Windows 10 won't accept them for its display, Not that it matters, mpv is quite happy with them.

 

Through my adventures, it is clear that HDR being handled by your display as opposed to tone mapping to the desktop, is an entirely different animal. Even though Windows is in HDR10, that peak luminance isn't fully accessible when tone mapping to the desktop. Windows just isn't allowing it. It's got a tight grip on it's desktop. Maybe in the future an update might loosen it's grip. That said, mpv is providing a very good picture, while not being as sparkly as HDR passthrough. Which is achievable when using jeeb's test build. Once the HDR only option for external players is implemented, I will test the bashiznits out of it :) 

 

I hope this is informative :)


  • PrincessClevage and markyp like this

#272 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 15 September 2019 - 08:42 PM

Just found this in the docs. I didn't use this setting. Let's see what it does :)

 

5d7eda3883752_targetpeak.jpg



#273 Jdiesel OFFLINE  

Jdiesel

    Advanced Member

  • Members
  • 2682 posts
  • Local time: 01:13 AM
  • LocationRegina, SK

Posted 15 September 2019 - 09:04 PM

This is the guide I used. ChromaPure isn't free but it does a good job and I've calibrated multiple displays with it and it has paid for itself many times over.

http://www.curtpalme...pic.php?t=35322

You can also use DisplayCal to generate a 3dlut to use once your display is calibrated. I don't know if mpv supports 3dluts yet but with madVR they can be used.
  • Doofus likes this

#274 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 15 September 2019 - 10:48 PM

Scratch that. I should have checked the logs :/


Edited by Doofus, 15 September 2019 - 11:44 PM.


#275 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 18 September 2019 - 02:13 AM

Now I've got it! I was over thinking it. Just had to set a few parameters, use a large pallet of 778 patches and let it do the rest. One important detail I discovered, was to not apply the ICC to the desktop, as well as apply it to mpv. You choose the option to use interactive display adjustment, under calibration, then it makes the ICC based on that and its measurements. So when mpv tonemaps, it will be in accordance with the calibration. If you apply it to the desktop, it in effect, is adjusting twice. And it looks like crap. But with this latest ICC, I'm even closer to how it looks with passthrough. It still doesn't match the peak brightness, but it's closer than it was. 

 

The game is afoot! :)


Edited by Doofus, 18 September 2019 - 02:15 AM.

  • lorac likes this

#276 PrincessClevage OFFLINE  

PrincessClevage

    Advanced Member

  • Members
  • 1039 posts
  • Local time: 05:13 PM

Posted 18 September 2019 - 04:09 AM

Wouldn’t it be better to put effort into finding the best way to support displays that are constant true HDR10 WCG going forward and get the best possible picture using devices, software and configuration that can support the original intention? And in the end would this not produce superior results

#277 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 18 September 2019 - 10:31 AM

Wouldn’t it be better to put effort into finding the best way to support displays that are constant true HDR10 WCG going forward and get the best possible picture using devices, software and configuration that can support the original intention? And in the end would this not produce superior results


That's what the original post is for. But the display should be properly calibrated and measured. Knowing what your white and black points are and having that information in an icc does make a difference. Also, Windows in HDR10 seems to throw off the color balance of the desktop. I was visually adjusting mine, and the Adobe icc I found, seemed to be somewhat close to that. What I have now isn't massively better than what I had, but it is more accurate. I haven't tested it with bt.709 stuff, yet. mpv's algorithms are designed to 'enhance' the picture and not simply follow the metadata. Like pulling more detail out of dark areas. Using an icc is more true to metadata transfer. I'm not finished testing, yet. Now that my display color is adjusted, I also want to see what difference that will make with mpv's algorithms.

#278 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 18 September 2019 - 06:26 PM

If you want the best picture with the greatest of ease, just use jeeb's build as an external player. Run Windows in SDR and have your display calibrated for HDR passthrough.

#279 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 18 September 2019 - 06:26 PM

Double post

Edited by Doofus, 18 September 2019 - 06:27 PM.


#280 Doofus ONLINE  

Doofus

    Advanced Member

  • Members
  • 12290 posts
  • Local time: 12:13 AM

Posted 19 September 2019 - 10:36 PM

If anyone wants to try the color profile I've made, I'm attaching it. You'll want to use it with a target-peak between 220 and 300, I use 300. And use this config.

[HDR]
profile-desc=cond:p["video-params/primaries"]=="bt.2020"
hdr-compute-peak=no
tone-mapping=reinhard
tone-mapping-param=0.5
tone-mapping-desaturate=3
icc-profile=C:\Windows\System32\spool\drivers\color\My profile 3.icm
target-peak=300

Remember that your white point will likely not match mine, so you'll have varying results.

Attached Files






0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users