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 11th July 2015, 23:39   #31681  |  Link
Eyldebrandt
Registered User
 
Join Date: Oct 2012
Posts: 26
Quote:
Originally Posted by madshi View Post
I'll reply to all your posts later. For now I would just like to share a new test build, using a tweaked SuperRes algorithm:

http://madshi.net/madVR8816b.rar

SuperRes image quality should generally be slightly better than in the official v0.88.16 build, furthermore I've added an anti-ringing filter which should noticeably reduce ringing artifacts introduced by SuperRes. I hope that the anti-ringing filter will allow us to use algo 2, which is my favorite algorithm, but didn't work well due to the added ringing artifacts in v0.88.16. FWIW, if I had to decide on final SuperRes configuration right now, it would be this:

- algo = 2
- use alternative color space = off
- low: strength=0.5; passes=1
- medium: strength=1.0; passes=1
- high: strength=1.0; passes=2
- ultra: strength=1.0; passes=4

Would like to have your feedback about this. If you believe that the above settings are a totally bad idea, please let me know. Ideally with a small sample which shows that other settings work much better. Or if you like my settings, please let me know, too. Thanks!


Well, SuperRes is, again, quite better than the previous build.
But it creates a huge amount of aliasing.

Here is a sample.
that's the Gone Girl Blu-ray (which is what we can call a very high quality content) upscaled to 3440x1440 résolution in 2.39 ratio.
Chroma is NNEDI3 64, image Doubling is NNEDI3 64, downscaling is CR AR LL.

Look at the police car, the framing of the rear door.

No SuperRes, No Adaptative Sharpen


SuperRes high, No Adaptative Sharpen


SuperRes ultra, No Adaptative Sharpen


SuperRes ultra, Adaptative Sharpen 0.3


No SuperRes , Adaptative Sharpen 1.5 (ridiculous setting, just for science, though)

Last edited by Eyldebrandt; 12th July 2015 at 00:15.
Eyldebrandt is offline   Reply With Quote
Old 11th July 2015, 23:47   #31682  |  Link
Ver Greeneyes
Registered User
 
Join Date: May 2012
Posts: 445
Quote:
Originally Posted by madshi View Post
Ideally an upscaler would work magic.
Like this? (top is processed, bottom is original) Of course there are some problem areas, and there are way more examples where this neural network generates extremely weird things, but I thought its performance on this particular image was pretty incredible.

Last edited by Ver Greeneyes; 12th July 2015 at 00:17.
Ver Greeneyes is offline   Reply With Quote
Old 11th July 2015, 23:56   #31683  |  Link
Anime Viewer
Troubleshooter
 
Anime Viewer's Avatar
 
Join Date: Feb 2014
Posts: 333
Quote:
Originally Posted by Eyldebrandt View Post
Here is a sample.
that's the Gone Girl Blu-ray (which is what we can call a very high quality content) upscaled to 3440x1440 résolution in 2.39 ratio.
Chroma is NNEDI3 64, image Doubling is NNEDI3 64, downscaling is CR AR LL.

Look at the police car.

No SuperRes, No Adaptative Sharpen


SuperRes high, No Adaptative Sharpen


No SuperRes , Adaptative Sharpen 1.5 (ridiculous setting, just for science, though)
What part of the police car do you want us looking at? Given its a high quality sample (blu-ray) I wouldn't expect us to see much of a difference, and the "no SuperRes/no A Sharp" looks just as good to me as the SuperRes'd images. The Adaptative Sharpen 1.5 (no SuperRes) image looks pretty good to me. On the 1.5 image I can read the 59 on the license plate pretty well, but on all the other images the 59 is blurred.
Perhaps a settings that uses certain features (like SuperRes) on low resolution content, and others active on higher resolution content (like AS) would be a good idea. I know we can do this with rules pretty easy right now, but in the interest of simplifying things it might be something to look into down the road.
__________________
System specs: Sager NP9150 SE with i7-3630QM 2.40GHz, 16 GB RAM, 64-bit Windows 10 Pro, NVidia GTX 680M/Intel 4000 HD optimus dual GPU system. Video viewed on LG notebook screen and LG 3D passive TV.
Anime Viewer is offline   Reply With Quote
Old 12th July 2015, 00:06   #31684  |  Link
Eyldebrandt
Registered User
 
Join Date: Oct 2012
Posts: 26
The rear window of the car.
The sample with AS @ 1.5 is ugly as hell. Wax on trees, loss details on fines structures, luma pixels oversharpened. But, less aliasing than ultra SuperRes.

The best combo is not on those samples, uit's NNEDI3 + low set for AS.

It's obvious to say a feature can be good for one content and not for an other.
But, the point is, if SuperRes might be destructive on these type of content, what is he doing on poor quality content ?
To my opinion, i have more to gives to Mathias with this kind of sample, rarer than 480p to 1080p content. I can post hundred of samples, but, hey, obviously, 480p to WQHD or 4K will look shit w/e you've used. So i'll focus on those who can show important details.
i.e, i have a sample who shows the destructive action of SuperRes to pixels when bitrate is high and amount of details is huge. Step by step.

Last edited by Eyldebrandt; 12th July 2015 at 00:10.
Eyldebrandt is offline   Reply With Quote
Old 12th July 2015, 00:09   #31685  |  Link
Anime Viewer
Troubleshooter
 
Anime Viewer's Avatar
 
Join Date: Feb 2014
Posts: 333
Quote:
Originally Posted by Ver Greeneyes View Post
Like this? Of course there are some problem areas, and there are way more examples where this neural network generates extremely weird things, but I thought its performance on this particular image was pretty incredible.
Which is the original image (top?), and which is the effected image (bottom?). What is it supposed to be doing to the picture? To me it looks like the top picture has textures, and may be real life pictures collaged together. The bottom looks like the image was transformed into an animated version. They are quite different images when compared (the dog in the middle with the black dots around its mouth in the bottom image is quite different from the same dog nose and mouth area in the top image, as are the fur and eyes on many of the dogs). They both look like very nice images, but are also very different looking.
Given its Google their source probably wouldn't be to accessible for integration into software like madVR and MPDN (from a rights perspective).
__________________
System specs: Sager NP9150 SE with i7-3630QM 2.40GHz, 16 GB RAM, 64-bit Windows 10 Pro, NVidia GTX 680M/Intel 4000 HD optimus dual GPU system. Video viewed on LG notebook screen and LG 3D passive TV.
Anime Viewer is offline   Reply With Quote
Old 12th July 2015, 00:13   #31686  |  Link
XMonarchY
Registered User
 
Join Date: Jan 2014
Posts: 489
Are all 3 madVR sharpening methods considered to be superior to TV's built-in sharpening post-processing? Is there a way to test that somehow?
__________________
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
XMonarchY is offline   Reply With Quote
Old 12th July 2015, 00:14   #31687  |  Link
Ver Greeneyes
Registered User
 
Join Date: May 2012
Posts: 445
Quote:
Originally Posted by Anime Viewer View Post
Which is the original image (top?), and which is the effected image (bottom?).
The bottom image is the original (counterintuitive I know). Check out some of the other images linked from that twitter bot.

Quote:
Originally Posted by Anime Viewer View Post
Given its Google their source probably wouldn't be to accessible for integration into software like madVR and MPDN (from a rights perspective).
Yes, I don't think the Apache license is compatible. But the bigger issue is that it's just a remarkably good looking example* It really does a number on images that aren't composed of animal heads - it's not suitable at all as a general algorithm unless you want to make your videos look extremely psychedelic.

* although it does add an extra head onto the body of the big dog in the bottom left

Last edited by Ver Greeneyes; 12th July 2015 at 00:23.
Ver Greeneyes is offline   Reply With Quote
Old 12th July 2015, 00:27   #31688  |  Link
Anime Viewer
Troubleshooter
 
Anime Viewer's Avatar
 
Join Date: Feb 2014
Posts: 333
Quote:
Originally Posted by Eyldebrandt View Post
The rear window of the car.
The sample with AS @ 1.5 is ugly as hell. Wax on trees, loss details on fines structures, luma pixels oversharpened. But, less aliasing than ultra SuperRes.

The best combo is not on those samples, uit's NNEDI3 + low set for AS.
I was looking at the image in the forum re-sized size.
When I look at them opened in separate browser tabs at the full image resolution I still like the AS only image best. To me the back car window (including the bricks you see through it), the camera, the roof tiles, the bricks, the stone walkway, the light on top of the cop car, the writing on everything (the car, the house number, the fore mentioned license plate). The only thing that I think may look less real is the bush by the far right house window. Where the leaves on the plant look like they could be more from a painting than actual plant. (I still think the branching/trunk of the bush looks better in the AS only image compared to the others).
Our difference in opinion on what looks good/best is subjective and why its nice to have choices as to what effects we'd each want to have enabled or disabled given particular content.

Quote:
Originally Posted by Ver Greeneyes View Post
* although it does add an extra head onto the body of the big dog in the bottom left
haha, I missed the ghost dog in my initial looks at those images.


Quote:
Originally Posted by Ver Greeneyes View Post
The bottom image is the original (counterintuitive I know). Check out some of the other images linked from that twitter bot.

Yes, I don't think the Apache license is compatible. But the bigger issue is that it's just a remarkably good looking example* It really does a number on images that aren't composed of animal heads - it's not suitable at all as a general algorithm unless you want to make your videos look extremely psychedelic.
Looking at the dog images, and learning what the real vs modified image was made me think it could be interesting for use on anime/drawn characters to show what a real life counterpart to the characters might look like. Then I saw the other images, and how it seems to like to put disturbing looking eyes on all the other pictures I'm seeing on http://psychic-vr-lab.com/deepdream/ (If that is the area you wanted me to see the other pictures?) Like you said its probably not all that useful in madVR where we want to keep images relatively close to the look of the original source material.
__________________
System specs: Sager NP9150 SE with i7-3630QM 2.40GHz, 16 GB RAM, 64-bit Windows 10 Pro, NVidia GTX 680M/Intel 4000 HD optimus dual GPU system. Video viewed on LG notebook screen and LG 3D passive TV.

Last edited by Anime Viewer; 12th July 2015 at 00:43. Reason: merged posts
Anime Viewer is offline   Reply With Quote
Old 12th July 2015, 00:46   #31689  |  Link
tFWo
Registered User
 
Join Date: Apr 2011
Posts: 24
Rings of Saturn are gone. Rings of Uranus remain. (madshi don't hate me, mods please don't ban me )

AR filter helps a lot. Alternative color space (acs) is just bad. I prefer algo 1.

But tonight I found out that s-xbr100+ and SuperRes don't like each other. Something is seriously wrong when you use s-xbr with sharpness 100 or higher and SR.

720p source

sxbr100-nosr

http://i.imgur.com/3ujev1b.png

sxbr100-sr-pass1-str1-algo1-noacs (This is what I called dirt in the last post. SR actually blurs/darkens the numbers on the plate.)

http://i.imgur.com/yYiNCzc.png

sxbr100-sr-p1-s1-a1-acson

http://i.imgur.com/TaQ2uOw.png

nn64-sr-p1-s1-a1-noacs (no blurring with NNEDI3)

http://i.imgur.com/YYCEMrK.png

nn64-sr-p1-s1-a1-acson

http://i.imgur.com/TIbk0BD.png

nn64-sr-p1-s1-a2-acson aka Rings of... :P (imagine how this looked without the AR filter)

http://i.imgur.com/Y2a0xuR.png

Last edited by tFWo; 12th July 2015 at 01:53.
tFWo is offline   Reply With Quote
Old 12th July 2015, 00:56   #31690  |  Link
aufkrawall
Registered User
 
Join Date: Dec 2011
Posts: 1,708
Quote:
Originally Posted by madshi View Post
My thinking was that image enhancements should only be activated on demand if you happen to play a source which is overly soft. As such, it would still make sense to run image enhancements and upscaling refinement at the same time.

But all of this is still up for discussion. We're not there yet, though. I want to first reduce all the sharpening options as much as possible. Only afterwards we'll go back to see when to use image enhancements vs upscaling refinement, in combination or not etc...
Ok, I will come back on it.

Quote:
Originally Posted by madshi View Post
Yes, NNEDI3 is still king. Totally agree there. But it's also still very slow.
With Hawaii GPU, image doubling with 64 neurons can be done for 1080p30 via DirectCompute:
http://www.forum-3dcenter.org/vbulle...0&postcount=18


Quote:
Originally Posted by madshi View Post
I don't know. So you'd be fine with my suggested SuperRes settings?
Yes, I think the values are useful with the current alogorithm.
But I think Shiandow should take a look at the issue Eyldebrandt reported. Maybe there is a connection to why SuperRes seems to become much more aggressive with NNEDI3 compared to super-xbr?
aufkrawall is offline   Reply With Quote
Old 12th July 2015, 00:58   #31691  |  Link
MS-DOS
Registered User
 
Join Date: Sep 2012
Posts: 75
Quote:
Originally Posted by madshi View Post
Any obvious changes compared to the previous build?
Much less ringing and probably other artifacts. The difference is bigger when more passes are used.
MS-DOS is offline   Reply With Quote
Old 12th July 2015, 01:07   #31692  |  Link
leeperry
Kid for Today
 
Join Date: Aug 2004
Posts: 3,463
Quote:
Originally Posted by madshi View Post
Yes, but since then I've done some tests and found that I always clearly preferred HQ on. And it's technically, scientifically better, too. And the majority of users seemed to agree, too.

HQ off produces a different look, one that I personally find very artificial looking.
I've had time to read the few pages I missed while I was away and it would appear that the "majority of users" you would be mentioning is actually two guys in this thread, maybe three.

Either way, I fully agree that you are the captain onboard and that what matters most is your own satisfaction eventually. I'm also glad and extremely thankful for the fact that you were kind enough to share this gem of a VR.

I rest my case that I'm quite sure nobody uses the "don't use linear light for dithering" debug option anymore and I would happily trade it for "don't use HQ downscaling for SuperRes" if any possible please.

Quote:
Originally Posted by madshi View Post
You mean softness, not sharpness, right?

If there are compression artifacts, that should be handled by a different filter. Which sadly doesn't exist yet.
I did, my bad. The softness knobs does come in handy sometimes, especially as sxbr50 is a ringing feast so decreasing sharpness and setting softness to 0.01 or 0.02 hits the spot very nicely IME. The latter can also indeed be increased to hide compression artifacts very effectively

Quote:
Originally Posted by madshi View Post
Try adding some AdaptiveSharpen, that makes the image look more like HQ off.
I really like SR's effect on motion-blur, I'm not interested in running another sharpening filter on top but thanks for the advice.

All this said, I'm currently equally impressed by my audio and video rigs and it seems that .15 is the last version of mVR that allows SR softness, disabling HQ and allowing to change the SR sharpness and number of passes(0.65/3 for me). I'm totally cool with sticking with it, I wish sxbr also came with a knob so I could try my luck with 35-40(25 being too soft and 50 too strong) but anyway I'm not interested in mVR with SR having HQ forced and 3 presets only so I might very soon find myself with a lot more free time on my hands...and with no regret at that coz PQ is stellar

I'll remain on the lookout for a version of mVR that will support dimensions changing PS scripts then

Last edited by leeperry; 12th July 2015 at 01:31.
leeperry is offline   Reply With Quote
Old 12th July 2015, 01:18   #31693  |  Link
Anime Viewer
Troubleshooter
 
Anime Viewer's Avatar
 
Join Date: Feb 2014
Posts: 333
Quote:
Originally Posted by leeperry View Post

I rest my case that I'm quite sure nobody uses the "don't use linear light for dithering" debug option anymore and I would happily trade it for "don't use HQ downscaling for SuperRes" if any possible please.
Are you referring to having "don't use linear light for dithering" checked or unchecked? I keep it checked as I see a huge performance hit if I leave it unchecked, so if you're advocating having it unchecked and the option removed then I'm against that. If on the other hand you're advocating having it checked by default with the option to uncheck it removed then I'd be fine with that configuration.
__________________
System specs: Sager NP9150 SE with i7-3630QM 2.40GHz, 16 GB RAM, 64-bit Windows 10 Pro, NVidia GTX 680M/Intel 4000 HD optimus dual GPU system. Video viewed on LG notebook screen and LG 3D passive TV.

Last edited by Anime Viewer; 12th July 2015 at 01:22.
Anime Viewer is offline   Reply With Quote
Old 12th July 2015, 01:26   #31694  |  Link
leeperry
Kid for Today
 
Join Date: Aug 2004
Posts: 3,463
Quote:
Originally Posted by Anime Viewer View Post
Are you referring to having "don't use linear light for dithering" checked or unchecked?
I didn't realize it had any effect on performance, I think you would be surprised how much banding this option adds to any gray ramp.......to the point that enabling debanding in mVR and checking this option would be utterly counterproductive IMHO. It might very well also affect your judgment on other post-processing settings.

Fair enough, well there's still some room for an extra option otherwise ^^
leeperry is offline   Reply With Quote
Old 12th July 2015, 05:21   #31695  |  Link
Warner306
Registered User
 
Join Date: Dec 2014
Posts: 1,111
Quote:
Originally Posted by JarrettH View Post
When image doubling, does chroma upscaling happen before chroma doubling?

In that case, would the chain look like...

chroma > super-xbr (chroma upscaling) > super-xbr (chroma doubling)
If you are trying to understand the process from a beginner's perspective and use MPC, the following guide is a good beginner's guide:

http://forum.kodi.tv/showthread.php?...555#pid1843555

http://forum.kodi.tv/showthread.php?...097#pid1849097

Last edited by Warner306; 12th July 2015 at 05:46.
Warner306 is offline   Reply With Quote
Old 12th July 2015, 07:26   #31696  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,190
Quote:
Originally Posted by Anime Viewer View Post
When I look at them opened in separate browser tabs at the full image resolution I still like the AS only image best.
The 1.5 extreme setting? There's no other image that has AS only enabled. I hope you don't mean that one because it looks absolutely terrible, the contrast it adds to the image alone is enough to gag over.

Quote:
Originally Posted by Eyldebrandt View Post
Here is a sample.
that's the Gone Girl Blu-ray (which is what we can call a very high quality content) upscaled to 3440x1440 résolution in 2.39 ratio.
Chroma is NNEDI3 64, image Doubling is NNEDI3 64, downscaling is CR AR LL.
The SR high no AS image is extremely subtle with lots of very minute changes. There is no significant change here (back up by the image analysis I did) so I think this particular setting for this content is pretty much pointless.

The SR ultra no AS image offers something more substantial but still subtle, more edges are done stronger. I generally like what I see.

The SR ultra 0.3 AS image is actually binary identical to the above no AS image, so you must've goofed it because at the very least there should be dithering differences.

Last edited by ryrynz; 12th July 2015 at 08:10.
ryrynz is offline   Reply With Quote
Old 12th July 2015, 07:53   #31697  |  Link
QBhd
QB the Slayer
 
QBhd's Avatar
 
Join Date: Feb 2011
Location: Toronto
Posts: 462
Quote:
Originally Posted by Eyldebrandt View Post
Well, SuperRes is, again, quite better than the previous build.
But it creates a huge amount of aliasing.

Here is a sample.
that's the Gone Girl Blu-ray (which is what we can call a very high quality content) upscaled to 3440x1440 résolution in 2.39 ratio.
Chroma is NNEDI3 64, image Doubling is NNEDI3 64, downscaling is CR AR LL.

Look at the police car, the framing of the rear door.
Not sure what the rest of you are looking at (including the one who posted this) but the last image is stunning. I have been reading a lot of the posts lately, and it seems everyone wants a soft image. Soft=blurry.... I would love a crisp and clear image (with no artifacts, ringing or aliasing of course) And I must say this last image is just that.

QB
__________________
QBhd is offline   Reply With Quote
Old 12th July 2015, 08:17   #31698  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,190
Quote:
Originally Posted by QBhd View Post
And I must say this last image is just that.
I'm not going to get into a sharpening preference discussion, but there's only so far you can go sharpening an image before image quality degrades and I think No SR + AS 1.5 kinda passed it.
Not saying there aren't some obvious "improvements" but there's also a lot not to like with that setting, I'm sure I could find something that suited my tastes better whilst attaining similar sharpness (if I wanted that)

Last edited by ryrynz; 12th July 2015 at 08:22.
ryrynz is offline   Reply With Quote
Old 12th July 2015, 10:15   #31699  |  Link
Schwartz
Registered User
 
Join Date: Dec 2012
Posts: 44
Quote:
Originally Posted by madshi View Post
Ideally an upscaler would work magic. In real life we have to make do with what science has given us so far. If you want an algo which doesn't ring by itself, try Bilinear, Nearest Neighbor or Gaussian. Ouch. Pretty much everything else rings. Even NNEDI3 adds a bit of ringing in some situations. As Hyllian already mentioned, some ringing is sometimes beneficial, which makes the whole thing difficult. E.g. try some frequency burst test patterns without ringing. You'll get very bad results.
True. I've been digging into the subject a bit since yesterday and benchmarked some scalers. I ended up just running Mitchell. I do a lot of picture downsizing with Mitchell and it turns out that it's pretty damn good for video upsizing as well. Probably nothing for the 'sharper' crowd but I find it's a nice compromise.

I've never really seen NNEDI3 ring, but then again I don't see sxbr 50-75 ring all that much either. But all of my image doubling happens to SD content where it can 'do some work', a bit of smoothing is usually desired. Pixel-perfect encodes I leave alone.

I second the user who said he'd enjoy more sxbr options. 60-65 might be the sweet spot for me. Maybe an input field?

Last edited by Schwartz; 12th July 2015 at 10:18.
Schwartz is offline   Reply With Quote
Old 12th July 2015, 11:55   #31700  |  Link
SithUK
Registered User
 
Join Date: Dec 2001
Posts: 16
I've given up on using a 3D LUT for now. I've tried turning the LAV filter decoding to None (instead of DXVA native), and using the Bilinear chrome upscaler. I tried ticking all boxes in the "trade quality for performance" section. Unfortunately my average render speed sometimes exceeds 50ms on busy action scenes, with 40ms being the minimum for a 25fps film.

It looks like a 5 year old laptop (i5 2.4Ghz/4Gb/HD 4500) is too old tech to manage a 3D LUT.

Fortunately I've sourced a new lamp for my JVC X3 and the colours look ok, so I can live with the setup for now. Thanks for the help in trying to get a 3D LUT to work for me.

Quote:
Originally Posted by SithUK View Post
I've had a look and the Ati 540v that I have is basically a mobility hd 4550. Userbenchmark.com has it as comparable to an Intel hd2000.

Should I try selecting all the tick boxes in the trade performance for quality section? The default install has the first half of the options selected already.

The default chroma, up scaling, and down scaling options are already as suggested above. (Bicubic and lanczos)

I tried adjusting the 3d lut bit rate option on the trade performance for quality page to the lowest value of 6. I still had stuttering playback.

Is there anyway to divert the load from the gpu to the cpu?my processor is showing max 6%, and that is likely for only one of the 4 cores.

Is my hope of using a 3d lut with this laptop realistic? (Cpu is i5 m450 @ 2.4Ghz, 4Gb ram, Ati 540v gfx)
SithUK 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 22:00.


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