Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Hardware & Software > Software players

Reply
 
Thread Tools Search this Thread Display Modes
Old 1st January 2015, 20:46   #27941  |  Link
Fullmetal Encoder
Registered User
 
Join Date: Jan 2011
Posts: 107
Quote:
Originally Posted by e-t172 View Post
Because the best image quality is obtained with a gaussian point spread function, which is the ideal goal for an image upscaler. Fixed-pixel displays such as LCDs typically have square ("box") PSFs, which is the exact opposite and possibly the worst PSF there is. For these displays, upscaling to a higher resolution is likely to look better than native resolution (picture size being equal, of course), because you get less of the square PSF of the screen and more of the near-gaussian PSF of the upscaling filter. A counter-example is CRTs which have a naturally gaussian PSF.

That said, for this to make a difference, you need to sit close to the screen, otherwise visual acuity is not enough to resolve the difference in high-frequency spatial information.
Interesting, so, if I follow you correctly, you are saying that the higher pixel density of a 4K panel allows the panel to more accurately display the higher quality output of the upscaling filter due to the nature of the filter's near-gaussian PSF? So, essentially, the panel acts as a sort of lens which filters the image being displayed. And the quality of the image which reaches our eyes is limited by and can only be as good as is permitted by that filter. And the panel having a square PSF is the weakest link in the chain.

It should be possible then to come up with an objective way to gauge the relative benefit of moving from a smaller screen to a larger screen given that we know the screen sizes and resolutions. I'm thinking that a larger panel would negate the benefit of the higher resolution somewhat. Maybe taking the ratio of each panel's resolution to it's size in comparison would provide a better answer to XMonarchY's question.

XMonarchY it may be wise to take into consideration the nature of the display you linked (that is a PV panel and perhaps more susceptible to gamma shift) as well as how close you will be to the screen. I would expect the gamma shift to be worse the closer you are.
Fullmetal Encoder is offline   Reply With Quote
Old 1st January 2015, 20:57   #27942  |  Link
Shiandow
Registered User
 
Join Date: Dec 2013
Posts: 752
Quote:
Originally Posted by e-t172 View Post
Because the best image quality is obtained with a gaussian point spread function, which is the ideal goal for an image upscaler.
I'm not sure why you think a gaussian PSF is the ideal goal for an image upscaler. If that was the case then using a Gaussian kernel for upscaling would be ideal, but that tends to result in images which are quite blurry.

As far as I could find the article you linked didn't claim that Gaussian was best, just that it was better than a box filter.

Last edited by Shiandow; 1st January 2015 at 21:15.
Shiandow is online now   Reply With Quote
Old 1st January 2015, 21:06   #27943  |  Link
Fullmetal Encoder
Registered User
 
Join Date: Jan 2011
Posts: 107
Quote:
Originally Posted by Qaq View Post
Set a 1080 display to 720 res and start a 720 movie. Now set display back to 1080 and start a 720 movie upscaled to 1080. See?
And when I do, the 1:1 image looks superior to my eyes. That's what I'm trying to reconcile.
Fullmetal Encoder is offline   Reply With Quote
Old 1st January 2015, 21:26   #27944  |  Link
e-t172
Registered User
 
Join Date: Jan 2008
Posts: 567
Quote:
Originally Posted by Fullmetal Encoder View Post
It should be possible then to come up with an objective way to gauge the relative benefit of moving from a smaller screen to a larger screen given that we know the screen sizes and resolutions. I'm thinking that a larger panel would negate the benefit of the higher resolution somewhat. Maybe taking the ratio of each panel's resolution to it's size in comparison would provide a better answer to XMonarchY's question.
The absolute limit of human visual acuity is 0.5 minute of arc, though 1 minute of arc seems more realistic especially with the lower background luminance in a home theater scenario. If you sit far enough from the screen that the pixel size if less than 1 minute of arc (which is a typical home theater best practice), you won't be able to see the difference between a box PSF and a gaussian PSF, because the distance between the centers of the pixels will be less than your visual acuity can resolve.

For this reason, this stuff only makes a difference if you're sitting very close to the screen (especially in 4K!). Basically, if you sit too close to a box PSF screen (e.g. LCD at native resolution), it will appear pixelated; if you sit too close to a gaussian (or near-gaussian) PSF screen (e.g. 1080p upscaled to 4K), it will appear blurry. The latter is preferable to the former.

Quote:
Originally Posted by Shiandow View Post
I'm not sure why you think a gaussian PSF is the ideal goal for an image upscaler. If that was the case then using a Gaussian kernel for upscaling would be ideal, but that tends to result in images which are quite blurry.

As far as I could find the article you linked didn't claim that Gaussian was best, just that it was better than a box filter.
True. Trying to determine the best PSF is akin to trying to determine the best upscaling filter, which often results in heated discussions There's no debate about box being the worst though. If I'm not mistaken it's mathematically equivalent to nearest neighbor upscaling.

Quote:
Originally Posted by Fullmetal Encoder View Post
And when I do, the 1:1 image looks superior to my eyes. That's what I'm trying to reconcile.
I'm not sure what you're doing, but if you want to compare 1:1 to upscaled in a meaningful way, you need to keep the image size the same. Which is impossible to do unless you have two similarly sized screens with different native resolutions. Or maybe you could try using one screen and moving back and forth to keep apparent size the same, but that doesn't sound very scientific!
e-t172 is offline   Reply With Quote
Old 1st January 2015, 22:31   #27945  |  Link
Qaq
AV heretic
 
Join Date: Nov 2009
Posts: 422
Quote:
Originally Posted by Fullmetal Encoder View Post
And when I do, the 1:1 image looks superior to my eyes. That's what I'm trying to reconcile.
I see. From your viewing distance you don't see bigger pixels on 720. And there are also no upscaling artifacts. Thats fine. But what about lower resolution? You won't set your display to 540, will you?
Personally I hate upscaling artifacts myself. But I can't stand to bigger pixels on 720 display even more. Thats why I always prefer 1080 if possible, even for WEB-DLs. But if I can't get 1080, I'm fine with upscale.
Qaq is offline   Reply With Quote
Old 1st January 2015, 22:55   #27946  |  Link
Fullmetal Encoder
Registered User
 
Join Date: Jan 2011
Posts: 107
Quote:
Originally Posted by e-t172 View Post
The absolute limit of human visual acuity is 0.5 minute of arc, though 1 minute of arc seems more realistic especially with the lower background luminance in a home theater scenario. If you sit far enough from the screen that the pixel size if less than 1 minute of arc (which is a typical home theater best practice), you won't be able to see the difference between a box PSF and a gaussian PSF, because the distance between the centers of the pixels will be less than your visual acuity can resolve.

For this reason, this stuff only makes a difference if you're sitting very close to the screen (especially in 4K!). Basically, if you sit too close to a box PSF screen (e.g. LCD at native resolution), it will appear pixelated; if you sit too close to a gaussian (or near-gaussian) PSF screen (e.g. 1080p upscaled to 4K), it will appear blurry. The latter is preferable to the former.
Yes, but therein lies the rub. If I get 40 inch 4K panel and have to sit so far away from it that the blurriness isn't an issue, then will it be bigger in my field of view (and better) than a smaller panel would be if I were to sit up close to it? If not then where is the benefit in higher resolutions and larger screens? Just buy a really high quality, color accurate 24-25" panel with good uniformity and sit right up close to it. Though all panel sizes being equal I can understand the benefit of higher resolution.

Quote:
Originally Posted by e-t172 View Post
I'm not sure what you're doing, but if you want to compare 1:1 to upscaled in a meaningful way, you need to keep the image size the same. Which is impossible to do unless you have two similarly sized screens with different native resolutions. Or maybe you could try using one screen and moving back and forth to keep apparent size the same, but that doesn't sound very scientific!
Yes, the best I can do is set my player to 1:1 then watch it upscaled.
Fullmetal Encoder is offline   Reply With Quote
Old 2nd January 2015, 08:29   #27947  |  Link
e-t172
Registered User
 
Join Date: Jan 2008
Posts: 567
Quote:
Originally Posted by Fullmetal Encoder View Post
Yes, but therein lies the rub. If I get 40 inch 4K panel and have to sit so far away from it that the blurriness isn't an issue, then will it be bigger in my field of view (and better) than a smaller panel would be if I were to sit up close to it?
Since visual acuity is expressed in angle units, it follows that the optimal field of view depends only on resolution, not screen size. This means the optimal field of view for 1080p is 32, and for UHD (~4K) it's - no surprise there - 64. You can watch 1080p on a >32 field of view if you upscale it on a 4K monitor; you won't get any more detail, but at least it won't look pixelated. If you do the same on a 1080p native monitor with a box PSF, then you won't get any more detail and it will look pixelated.

Quote:
Originally Posted by Fullmetal Encoder View Post
If not then where is the benefit in higher resolutions and larger screens?
The benefit of higher resolutions (screen size being equal) is, you get a more detailed and less pixelated image, assuming you sit close enough to the screen for it to make a difference (~1 MOA/pixel).

The benefit of larger screens (resolution being equal) is, you get to sit farther away from the screen and still enjoy the same level of detail.

Last edited by e-t172; 2nd January 2015 at 08:31.
e-t172 is offline   Reply With Quote
Old 2nd January 2015, 19:59   #27948  |  Link
XMonarchY
Registered User
 
Join Date: Jan 2014
Posts: 487
I am very sorry if this has been re-posted a dozen times, but I have been out of 3DLUT-making scene for a while now. I just made a madVR 3DLUT with dispcalGUI 2.6.0.0, but HCFR does not detect madVR. I uninstalled madVR using the included .bat in normal mode and As Administrator. Then I restarted. Then I placed madVR on root directory of drive D to prevent any long directory names with spaces and ran the install .bat file in normal mode and As Administrator. However, none of that worked. When I select madVR as a pattern generator in the latest HCFR 3.1.6, I get an error saying that madVR DLL was not found... I have the latest madVR x.x.11. Maybe all these new connection options need to be used, but I do not see any options for them. I do see my local IP in madTPG - 192.168.1.96. When my Windows 8.1 firewall asked me about allowing madTPG access, I clicked "Allow" for both public and private networks. I also removed and re-installed HCFR Calibration several times in normal and As Administrator, but that didn't help... It keeps telling me DLL was not found and asking me whether madVR was installed...

Please don't yell at me for asking, but how do I fix this problem? I figured the problem must be with HCFR because dispcalGUI 2.6.0.0 detected madTPG just fine when Profiling (reading patterns) for 3DLUT creation. I applied 3DLUT and I can definitely see image differences, which means that 3DLUT is working, but I need to verify its accuracy with HCFR (or is there some other way?). I can't post on AVS forums because they banned me (yes SHAME on me, its embarrassing and I regret it and still feel guilty, please don't rub it n... ) and I tried 10x times to write them e-mails to get back, but it didn't work and they catch IP's if you try to create another account, etc. Please help!
__________________
8700K @ 5Ghz | ASUS Z370 Hero X | Corsair 16GB @ 3200Mhz | RTX 2080 Ti @ 2100Mhz | Samsung 970 NVMe 250GB | WD Black 2TB | Corsair AX 850W | LG 32GK850G-B @ 165Hz | Xonar DGX | Windows 10 LTSC 1809

Last edited by XMonarchY; 2nd January 2015 at 20:12.
XMonarchY is offline   Reply With Quote
Old 2nd January 2015, 22:29   #27949  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 5,657
try version 86.10 not 86.11.

and dispcalGUI can this to. just use verify calibration.
huhn is offline   Reply With Quote
Old 3rd January 2015, 07:46   #27950  |  Link
Arm3nian
Registered User
 
Join Date: Jul 2014
Location: Las Vegas
Posts: 177
Does madVR do more work if using SVP to go from 24>60fps? And would converting 24 to 60 mean render times would need to be under 16 seconds instead of 41?
Arm3nian is offline   Reply With Quote
Old 3rd January 2015, 08:29   #27951  |  Link
Zachs
Suptitle, MediaPlayer.NET
 
Join Date: Nov 2001
Posts: 1,721
Quote:
Originally Posted by Arm3nian View Post
Does madVR do more work if using SVP to go from 24>60fps? And would converting 24 to 60 mean render times would need to be under 16 seconds instead of 41?
Yes that's right.
Zachs is offline   Reply With Quote
Old 3rd January 2015, 14:46   #27952  |  Link
jdl
Registered User
 
Join Date: Jun 2003
Posts: 12
Quote:
Originally Posted by Ceremony View Post
What are you using? Im going with an android smartphone for controlling it.
Yatse
jdl is offline   Reply With Quote
Old 4th January 2015, 21:22   #27953  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,137
Quote:
Originally Posted by GCRaistlin View Post
madVR seems to be very unstable when playing videos on the 2nd monitor. Though MPC-HC is unstable, too (crashes with EVR CP), but sometimes I get madVR debug window (with "Close application", "Continue" buttons). I remember you confirmed there are issues with the palyback on the 2nd monitor but the main advice was to playback videos on the 1st monitor. Is there any sense to try to reproduce the steps that lead to crash to report it to you, to send you any info regarding the 2nd monitor playback issues?
Many people are doing playback on 2nd monitor without any (major) problems. There are certain issues, like focus getting lost, but there should be no crashes. If you get crashes from madVR (with the "Close application" etc buttons), please always do send the bug reports. But if MPC-HC crashes with EVR, too, it's going to be hard to track down who or what is ultimately responsible.

Quote:
Originally Posted by DigitalLF View Post
Hello! i really love MadVR it just gets better and better! but one thing is missing! a seekbar inside the picture area like in a 2.35:1 movie. Why i wonder is because i'm about to order a 2.40:1 screen!
It's on my to do list.

Quote:
Originally Posted by x7007 View Post
NEEDI3 chroma upscaling + 970

doesn't work properly

Asus Strix 970
344.75

and there are so many bugs trying to watch MKV movies. black flashes, video is continuing rendering subtitles when paused and sometimes the video freeze and audio continues, sometimes it just doesn't continue at all and freeze on the last paused image, and sometimes it just freeze randomly when fast forward or back forward. and sometimes just low performance the renderer is doing 1-7 intead 6-7, and it just dropping frames non stop.

Using Potplayer + MADVR + LavFilters CUVID

So many video bugs I can't watch a movie properly. to fix the video I need to reopen it stop it or switch to JINC and it fix the slow down instant.

With 7970 same settings I didn't have a single issue
If the same thing worked fine with the 7970 then it sounds like an issue either with your 970 or with the 970 drivers. Of course it could also be a bug in madVR, but since it seems to work fine for many other people I wouldn't know what to look for. Generally fixing bugs is hard (or even next to impossible) if they only occur for one user.

Quote:
Originally Posted by yesgrey View Post
Is it possible to use madVR as a video processing filter?
Not at this time.

Quote:
Originally Posted by Xello View Post
Is there a person who can analyze madVR bug reports or guide me in analyzing my own to nail down the cause of a crash ?
Yes, me, of course.

Quote:
Originally Posted by dbcooper View Post
With the new version of madVR, I can't force film mode for a DVD. Weird.
What happens if you try? Your computer explodes? Some more details would be helpful. Very helpful would also be if you tried older madVR version to find with which version exactly the feature broke for you (if madVR is responsible, that is).

Quote:
Originally Posted by Neet009 View Post
madVR version 0.87.11
I want to use HCFR(version 3.1.6) and madTPG to make a display measure, but it showing this message to me [...]
Sorry, that's a bug in v0.87.11, it will be fixed in v0.87.12.

Quote:
Originally Posted by MSL_DK View Post
It is not madVRs fault. HCFR is not updated to version 0.87.11
madVR/madTPG is supposed to be backwards compatible to older calibration software versions. So it's my fault. Which is good, because I can fix it easily.

Quote:
Originally Posted by yok833 View Post
Dear Santa Claus Madshi,

We have all been very nice this year... so will you bring us presents for christmas ??
Hmmmm... Yes, there's a good chance there might a new version out with some nice new features in time for xmas this year...

Quote:
Originally Posted by Asmodian View Post
Without the "delay" options checked the queue sizes have no effect on seeking delays. madVR does not wait for them to fill but presents the first frame as soon as it is available.
Yes and no. When pre-presenting multiple frames in advance, after a seek madVR has to wait until all pre-presented frames were displayed, before new frames can be presented. So the number of pre-presented frames can have an influence on the seek times.

Quote:
Originally Posted by zerowalker View Post
In an update to the slow seeking i have.

I notice that if the content is MPEG-2 it's fast like EVR, but with H264 it's much slower.
That's weird. Is this independent on the decoder? And independent on software vs. hardware decoding etc?

Quote:
Originally Posted by digitech View Post
Anyone have recommended settings for sd content using neddi, i have no clue what combination to use, 90% of my sd content is animation and i love to use SVP with it, im on nvidia gtx 760, Intel i5 4670k, 16 gb ram, win7
If you're short on GPU power, use NNEDI3 only for luma doubling, not for anything else. Can't say anything about SVP, I don't have experience with SVP.

Quote:
Originally Posted by braddock View Post
I'm getting dropped frames every few minutes on a 1080p movie that's 4:2:2 instead of the usual blu-ray 4:2:0

GTX 750ti
i5 3570k
16gb ram

Is my GPU?
In the debug OSD (Ctrl+J), wenn the drops occur, are any of the queues near empty? If so, which ones exactly?

Quote:
Originally Posted by a8213711 View Post
My GPU is capable of i25 (see http://forum.doom9.org/showpost.php?...ostcount=27887 for details), is there a way to understand if my TV is capable too?
i25 should be the same as i50, which most TVs do support. But I would stay away from anything interlaced unless oyu have no other choice.

Quote:
Originally Posted by Anima123 View Post
I remember there's a way to modify the registry to force madVR to use nVidia GPU with Optimus on some notebooks, but I cannot find it anymore.
That only works for OpenCL, though, not for Direct3D.

Quote:
Originally Posted by Fullmetal Encoder View Post
How could an upscaled image look better than the original source material at native resolution? Every time I upscale a lower resolution source to 1080 on my monitor it looks worse with at least a loss in sharpness.
It's not a fair comparison to look at an image unscaled and then to scale it up and look at it from the same distance. The real question is whether a 1080p movie looks better on a 4K display or on a 1080p display, when both displays have the same size, and when you watch both display from the same distance. You can't take a shortcut and compare two images at different sizes. That just won't give you a proper answer to the real question.

To get an answer to the real question, read my comment here:

http://www.avsforum.com/forum/24-dig...l#post28854034
madshi is offline   Reply With Quote
Old 4th January 2015, 21:27   #27954  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,137
madVR v0.87.12 released

http://madshi.net/madVR.zip

Code:
* fixed: v0.87.11 broke madTPG API compatability
* madTPG: added dropdown box to allow manual gamut/3dlut selection
* madTPG: loading a 3dlut now takes immediate effect in madTPG
* madTPG: added API "GetVersion"
* madTPG: added API "Get/SetSelected3dlut"
* madTPG: number of pre-presented frames is now limited to max 3
Sorry, still only madTPG changes.
madshi is offline   Reply With Quote
Old 5th January 2015, 04:43   #27955  |  Link
James Freeman
Registered User
 
Join Date: Sep 2013
Posts: 919
You have absolutely nothing to be sorry about.
Thank you madshi for your time and devotion.
__________________
System: i7 3770K, GTX660, Win7 64bit, Panasonic ST60, Dell U2410.
James Freeman is offline   Reply With Quote
Old 5th January 2015, 09:57   #27956  |  Link
StinDaWg
Registered User
 
Join Date: Jan 2014
Posts: 216
I've had an issue for awhile now where I will play a video in FSE mode and the render time will settle down to say, 30ms for 720p->1080p NNEDI3 upscaling. Then I'll back out to windowed mode and back into FSE and the render time will drop down to a steady 25ms. I'm not sure what's happening but I can reproduce this consistently. I have my gpu overclocked and it's almost like madVR isn't accepting the overclock in FSE even though Afterburner says it's using 1050 core the entire time. No change with v0.87.12. I'm using an AMD 7850 and it happens on all firmware versions.

Does this happen to anyone else?
StinDaWg is offline   Reply With Quote
Old 5th January 2015, 10:24   #27957  |  Link
DarkSpace
Registered User
 
Join Date: Oct 2011
Posts: 204
Quote:
Originally Posted by madshi View Post
Sorry, still only madTPG changes.
As James Freeman noted, don't worry, you're doing this for free, after all.
Also, you did (almost) promise us a new version already, after all...
Quote:
Originally Posted by madshi View Post
Hmmmm... Yes, there's a good chance there might a new version out with some nice new features in time for xmas this year...
DarkSpace is offline   Reply With Quote
Old 5th January 2015, 10:39   #27958  |  Link
Neet009
Registered User
 
Join Date: Jan 2014
Posts: 21
Quote:
Originally Posted by madshi View Post
madVR v0.87.12 released
Thank you for releasing the new version. Now it works well with HCFR.
Neet009 is offline   Reply With Quote
Old 5th January 2015, 18:27   #27959  |  Link
shaolin95
Registered User
 
Join Date: Aug 2005
Posts: 241
Hello guys!
Is it normal for the Nvidia Control Panel settings like deinterlace etc to have no effect on the image when using madvr? I have a GTX750ti and Windows 8.1

Thanks
shaolin95 is offline   Reply With Quote
Old 5th January 2015, 19:50   #27960  |  Link
leeperry
Kid for Today
 
Join Date: Aug 2004
Posts: 3,462
Quote:
Originally Posted by shaolin95 View Post
Is it normal for the Nvidia Control Panel settings like deinterlace etc to have no effect on the image when using madvr?
yep, but you can always try DXVA and CUVID:

Quote:
Originally Posted by madshi View Post
features: (..)
- bypasses graphics card's video (damage) algorithms
leeperry is offline   Reply With Quote
Reply

Tags
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 14:57.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2019, vBulletin Solutions Inc.