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 22nd February 2016, 05:41   #36361  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,127
Quote:
Originally Posted by XRyche View Post
Is there any way to set up a profile based on what resizer is used? For example: If you only wanted to use a post processing sharpener if Jinc AR was used as an upscaler.
kind of complicated but yes this is possible.
huhn is offline   Reply With Quote
Old 22nd February 2016, 05:43   #36362  |  Link
XRyche
Registered User
 
Join Date: May 2008
Posts: 211
Quote:
Originally Posted by huhn View Post
kind of complicated but yes this is possible.
Any idea how I would go about it because I'm stumped?
__________________
Intel i5 3470, EVGA GTX 1050Ti SC ACX 2.0, Windows 10 Pro 64 bit, 16 GB 1600 mhz DDR3 RAM
XRyche is offline   Reply With Quote
Old 22nd February 2016, 06:06   #36363  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,127
Quote:
Originally Posted by XRyche View Post
Any idea how I would go about it because I'm stumped?
for of all you need a profile that is using jinc3AR when the conditions are meet and this profile is using the correct image refinement.

so the first step is creating profiles that are jinc when you want to use jinc.

so you are not setting up a profile depending on the resizer you make sure the correct resizer is used with the correct image refinement.

same result different approach.
huhn is offline   Reply With Quote
Old 22nd February 2016, 06:09   #36364  |  Link
Warner306
Registered User
 
Join Date: Dec 2014
Posts: 1,127
Quote:
Originally Posted by huhn View Post
"better" is clearly not the term i would use here "different" is more like it.

it is clearly against the creators intent and a good source doesn't need sharpening.
I've embraced image sharpening as an enhancement of the content material.

The content creator did not plan for poor lighting conditions, bad captures, source compression or many other things that go into producing a Blu-ray or broadcast TV series. Image sharpening can bring out detail that is in the source but is hard to see on a typical display without making it look unnatural.

However, I'd like to see what supersampling looks like compared to applying light sharpening to the unscaled source.
Warner306 is offline   Reply With Quote
Old 22nd February 2016, 06:30   #36365  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,481
Quote:
Originally Posted by huhn View Post
and they have a vivid mode too so what's the point?

it's not mastered to be watched this way.

whatever someone does with the image is up to the user. if they want to watch it the correct way or in vivid mode is there choice.
The point is the vast majority of the public perceive a sharper brighter picture to be better. Sharpness to a degree is perceived resolution enhancement and it does indeed trick the eye quite nicely in doing this.
ryrynz is offline   Reply With Quote
Old 22nd February 2016, 06:39   #36366  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,127
if you want to trick your self fine by me.

if someone want to use very high brightness and turn people into plastic there choice.

i never said something like this should be removed.

upscale refinement make by the way a lot of sense even with an good source. because you are usually loosing sharpness when you are scaling.
huhn is offline   Reply With Quote
Old 22nd February 2016, 06:49   #36367  |  Link
XRyche
Registered User
 
Join Date: May 2008
Posts: 211
Quote:
Originally Posted by huhn View Post
for of all you need a profile that is using jinc3AR when the conditions are meet and this profile is using the correct image refinement.

so the first step is creating profiles that are jinc when you want to use jinc.

so you are not setting up a profile depending on the resizer you make sure the correct resizer is used with the correct image refinement.

same result different approach.
I understand this approach and have actually been doing just that but using exact resolutions. I was looking for a way to make it more general by maybe specifying the use of the resizer.

Thanks for your suggestion huhn.

It's too bad resizer use isn't included in the scripting language.
__________________
Intel i5 3470, EVGA GTX 1050Ti SC ACX 2.0, Windows 10 Pro 64 bit, 16 GB 1600 mhz DDR3 RAM
XRyche is offline   Reply With Quote
Old 22nd February 2016, 06:58   #36368  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,481
Quote:
Originally Posted by madshi View Post
Yes, in this specific test image super-xbr 150 without AR looks better. But there are other tests images where it's the opposite, e.g. see aufkrawall's favorite Anime test video.
I don't think the Family Guy video is a particularly good test image for chroma. The only major point of difference is the stars in the headband which is very flat in color and has very little surrounding elements. In an example Jinc offered overly bright colors in comparison to super-xbr which he aufkrawall liked but actually wasn't accurate at all, you mentioned it was beneficial but to me it's just overly bright and not really accurate.

Your chromashort from Samsara (Jame's third image linked a couple of pages back) is another good test, here it appears that super-xbr 150 is too sharp as the red edges of the robes stand out far too much. 100 seems to be a better fit here, but I'm considering Reconstruction soft possibly with SuperRes too. Both of these seem to be better all purpose fits.

Quote:
Originally Posted by huhn View Post
if you want to trick your self fine by me.

if someone want to use very high brightness and turn people into plastic there choice.

i never said something like this should be removed.

upscale refinement make by the way a lot of sense even with an good source. because you are usually loosing sharpness when you are scaling.
Have you even done any supersampling tests?

You said yourself not required if the content is good.

Thing is not all content is good, or people think it's looks good. You can't match exactly what a native 1080 image would look like anyway
So if you're upscaling at all, you're basically tricking yourself...

Also there's content out there which are just 720 upscales in 1080, which could possibly benefit here..

Perhaps you could comment after doing tests, rather than blindly dismissing the benefits.

Quote:
Originally Posted by chros View Post
Interesting, thanks. Can you make 1 more test with the same image using Jinc/AR? Thanks
Jinc vs AR

For anyone interested based on what I've seen so far, super-xbr 100 (non AR) seems to provide me with the most natural looking results
and that'll be my default going forward until something looks objectively better (which will quite possibly be Reconstruction)

Last edited by ryrynz; 22nd February 2016 at 11:23. Reason: Mistake.
ryrynz is offline   Reply With Quote
Old 22nd February 2016, 07:46   #36369  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,127
Quote:
Originally Posted by ryrynz View Post
Have you even done any supersampling tests?

You said yourself not required if the content is good.

Thing is not all content is good, or people think it's looks good. You can't match exactly what a native 1080 image would look like anyway
So if you're upscaling at all, you're basically tricking yourself...

Also there's content out there which are just 720 upscales in 1080, which could possibly benefit here..

Perhaps you could comment after doing tests, rather than blindly dismissing the benefits.
you may read this again:

http://forum.doom9.org/showpost.php?...ostcount=36361 i haven't changed it at all.

so what's the problem?

and remember your old answer.

and most important read this question again.
huhn is offline   Reply With Quote
Old 22nd February 2016, 08:34   #36370  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,481
Quote:
Originally Posted by huhn View Post
you may read this again:

http://forum.doom9.org/showpost.php?...ostcount=36361 i haven't changed it at all.

so what's the problem?

and remember your old answer.

and most important read this question again.
I'll let this speak for itself.
This is a "good" quality source @ 1080 and the benefits are easily noticeable for me.
Look at the jaggies. Poof.. and this is only at 64 neurons..

Not just different.. this is an improvement.

Quote:
Originally Posted by Warner306 View Post
The content creator did not plan for poor lighting conditions, bad captures, source compression or many other things that go into producing a Blu-ray or broadcast TV series.
Are you always watching straight from blu-ray or playing full rips? Something to factor in as well..

It's fairly safe to say that no matter how "good" the quality of the full HD video, there's possibly something to be gained from image enhancements.
Whether that's just intelligent line construction, thinning or perceptual sharpness.. Native 1080 is not as good as it gets.
And that's the whole reason behind this feature. The benefits can be reasonably small, but your 1080 content won't look better.
I guess the purists will just miss out.

Last edited by ryrynz; 22nd February 2016 at 09:02.
ryrynz is offline   Reply With Quote
Old 22nd February 2016, 09:12   #36371  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,127
and the aliasing is just there because it is a terrible source.
i never said it is useless just a different approach to change the image.
huhn is offline   Reply With Quote
Old 22nd February 2016, 09:34   #36372  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,481
Quote:
Originally Posted by huhn View Post
and the aliasing is just there because it is a terrible source.
i never said it is useless just a different approach to change the image.
I guess you only play master copies then? I'm proving the point that even reasonably high quality 1080 content that not everyone is playing back isn't exactly perfect. I guess nothing is "good" unless you say it is..

Yeah, all we're doing is just changing the image

I'll leave it at that.
ryrynz is offline   Reply With Quote
Old 22nd February 2016, 09:46   #36373  |  Link
Thunderbolt8
Registered User
 
Join Date: Sep 2006
Posts: 2,188
Quote:
Originally Posted by huhn View Post
"better" is clearly not the term i would use here "different" is more like it.

it is clearly against the creators intent and a good source doesn't need sharpening.
is it just for sharpening or primarily to avoid chroma upsampling?
__________________
Laptop Lenovo Legion 5 17IMH05: i5-10300H, 16 GB Ram, NVIDIA GTX 1650 Ti (+ Intel UHD 630), Windows 10 x64, madVR (x64), MPC-HC (x64), LAV Filter (x64), XySubfilter (x64) (K-lite codec pack)
Thunderbolt8 is offline   Reply With Quote
Old 22nd February 2016, 10:10   #36374  |  Link
har3inger
Registered User
 
Join Date: Feb 2014
Posts: 139
There is definite anti-aliasing and line-thinning after double/halving in that sample, ryrynz. However, the source is also heavily aliased. In sources where there isn't heavy aliasing, doubling/halving probably just introduces unfaithfulness to the original image, to questionable benefit. You can already see the bit of white pant leg by the coffee table is thicker after doubling/halving. While it's a tiny detail that's worth the antialiasing effect, it's a pretty large deviation from what the source shapes were originally. But hey, if the option's there, and there exist bad 1080p sources to be played, the feature remains very useful.

I also like using SXBR without AR for chroma, although I don't like it as high as 150. Without question, the AR filter for SXBR chroma rings like crazy, and perhaps rings more than even lanczos 4 no AR. IMO, any significant ringing introduced into chroma is flat out unacceptable, as it presents as negative-color (green for red, yellow for blue, etc) halos that end up being super ugly. Turning off the AR filter for SXBR removes almost all ringing but loses you some source chroma details in surfaces by flattening them out, which can either be beneficial (like softcubic can be), or unnoticed, since most textures on surfaces are actually in the luma or duplicated there in full resolution.

150 is a bit too high for my taste. While it's super sharp and looks closest to 4:4:4, it accomplishes this with a side effect of making dark lines darker, bright lines brighter, and generally most lines thicker. This can cause some color shifts and increased saturation, and in some cases, color bleed. These problems start going away at SXBR 100 and are pretty much nonexistent at SXBR75. For sure, some samples look fantastic with 150 sharpness, but generally it's not that safe to try and recreate sharp 4:4:4 chroma from 4:2:0 with no luma information to inform the upscaler.

I've personally settled on sxbr 75 no AR for chroma upscaling. It's not too soft, never aliases, and doesn't ring. In most samples it's almost the same as nnedi256, but costs even less than jinc3 AR. It's clearly better than any of the bicubic scalers, so it's an ideal choice for me.
har3inger is offline   Reply With Quote
Old 22nd February 2016, 10:11   #36375  |  Link
huhn
Registered User
 
Join Date: Oct 2012
Posts: 7,127
Quote:
Originally Posted by Thunderbolt8 View Post
is it just for sharpening or primarily to avoid chroma upsampling?
chroma subsampling is not done by madVR and the output resolution doesn't change with supersampling. the filters are just used at a higher resolution.

for example scaling can work as a anti aliasing filter something madVR couldn't do before.
huhn is offline   Reply With Quote
Old 22nd February 2016, 10:23   #36376  |  Link
har3inger
Registered User
 
Join Date: Feb 2014
Posts: 139
Quote:
Originally Posted by ryrynz View Post

Jinc vs .10 AR
Jinc vs .11 AR
Jinc .10 AR vs Jinc .11 AR Big improvement (I think leeperry will be happy bout this)

For anyone interested based on what I've seen so far, super-xbr 100 (non AR) seems to provide me with the most natural looking results
and that'll be my default going forward until something looks objectively better (which will quite possibly be Reconstruction)
Wait, was the AR filter for upscaling updated as well? Changelog just says it's for downscaling, but the results you show here are clearly different. Is this jinc downscaling?
har3inger is offline   Reply With Quote
Old 22nd February 2016, 10:23   #36377  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,137
Quote:
Originally Posted by Stereodude View Post
I only made one log, so it can't be the wrong one. I can repeat the test and make a log each way if that would help.

What are the dropped frames from? It drops a few frames at the very start of playback (not unexpectedly), but the drops after that are a mystery to me since the queues don't empty and the max rendering time is still way less than 1/refresh.
From what I remember I think it might have been due to your refresh rate being lower than the movie frame rate. But if that's the cause it should be totally independent of the debanding functionality.

Quote:
Originally Posted by Unr3aL View Post
I think this was discussed a long time ago. It's hard to say now, but wasn't the issue tied to multi-monitor setups alone?
I thought it had something to do with timings not playing well if one of the monitors had a high (120, 144Hz) refresh rate and happened to be the one the movie was played on...
It could've also just been a wild guess at that time.

I myself happen to have a 120Hz monitor and I use a projector for movies.
Playing @120Hz on the monitor never really worked for me without a lot of visible jerking and judder as well as tons of presentation glitches.
The only thing that helped was going back to old path and playing around with the number of backbuffers used.
That issue (for me) dates back to when the new render path was introduced.

I've never tried upping the GPU queue though, so far I am auto-changing to 60Hz - minor annoyance...
Well, you could try a bigger GPU queue, maybe it helps.

Quote:
Originally Posted by RyuzakiL View Post
SOrry for the noob question but for what use is the scale in sigmoidal light in Image Upscaling?

I don't see any noticeable changes in PQ with it on but I notice some slight decrease in performance.
It performs scaling in a different "light space". It's not linear light, but something similar. In some situations it helps reducing aliasing a bit. Sometimes it can also be slightly detrimental. So it's your choice whether to use it or not.

Quote:
Originally Posted by RyuzakiL View Post
and also ctrl+j reports that I'm still using Lanzcos3AR even if I selected JincAR as my Image Upscaler so what gives??
There is probably a good reason for that. Maybe I can see why if you post a screenshot of the OSD with v0.90.11 (or newer).

Quote:
Originally Posted by huhn View Post
even 4.0 sharp edges or crispen edge has nearly no effect when scaling 1080p to 2160p even with disabled AR filter.
That will depend on the source image. The softer the source the less effect sharpening has, because sharpening tries to work out fine detail or edges. If there is no fine detail and no clear edges then sharpeners don't have a lot to work with. Add to that that sharpening in supersampled resolution is less effective than sharpening in the original resolution.

Quote:
Originally Posted by CarlosCaco View Post
Hey madshi i tried to test supersampling on 720p content on my 1360x768 screen, the doubling is working but upscaling refinements not
The current build has a special code path for when the source video has the same resolution as the screen. In your case the resolution differs, so that special supersampling code path isn't activated, sadly.

At some point I'm probably going to add a general supersampling option which is separate from doubling. I thought about adding it now, but couldn't find a good place for it. So I took the easy way out and simply renamed one of the doubling options and made it activate only if video res and screen res match. There'll be a solution for you in a future version, but I can't say when exactly.

Quote:
Originally Posted by Thunderbolt8 View Post
what does this mean? that normal BD playback on a 1080p screen now looks better with all this applied than when just viewing the movie normally?
The main purpose of supersampling is to improve the "look" and quality of sharpening.

Quote:
Originally Posted by leandronb View Post
Updated to the newer version and after a while with no problems it gave me the same error.
I used the debug mode to create the log, it created a 69MB file and i compressed it to 3mb and uploaded to mega.
Steps i did, opened an episode, went fullscreen, skipped 1, skipped 2 and it gave me the black screen i said earlier, hit esc to exit fullscreen and closed mpc.
this is the link:
https://mega.nz/#!tAIVEIoL!YjThiXXYo...2yv0isJBuGLuGU

thanks in advance.
Ok, thanks, will have a look when I find some time. Might have to wait until next weekend.

Quote:
Originally Posted by huhn View Post
"better" is clearly not the term i would use here "different" is more like it.

it is clearly against the creators intent and a good source doesn't need sharpening.
I understand your sentiment, but I'm not sure I agree with it. If we had 8K displays and lossless 4:4:4 high bitdepth 8K scans of the original negatives I might tend to agree with you, but we don't. As it stands, the movie studio sometimes has a 4K master, sometimes a 2K master, and then applies a lot of processing to turn it into a Blu-Ray encoding. Resolution is reduced (for 4K masters), colors are reduced, bitdepth is reduced, chroma resolution is reduced, almost always some amount of DNR is applied, and finally starving bitrates turn sharp masters into soft encodings.

I've seen a couple of 4K masters, grabbed from Sony's 4K service and reencoded using low bitrate. Those masters were originally encoded by Sony with not a lot of bitrate, and then grabbed and reencoded. They are really not in a good shape. But still, downscaling those 4K masters to 1080p using a good madVR downscaling algo produces clearly sharper and more detailed results than the original 1080p Blu-Ray (which was created using a much higher quality of the same original 4K master). Which shows that Blu-Rays are generally a bit on the soft side, compared to how they *should* look like, given the original master.
madshi is offline   Reply With Quote
Old 22nd February 2016, 10:24   #36378  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,137
Quote:
Originally Posted by har3inger View Post
Wait, was the AR filter for upscaling updated as well? Changelog just says it's for downscaling, but the results you show here are clearly different. Is this jinc downscaling?
I wondered the same thing. Upscaling AR filter was not modified in any way.
madshi is offline   Reply With Quote
Old 22nd February 2016, 10:40   #36379  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,137
madVR v0.90.12 released

http://madshi.net/madVR.zip

Code:
* fixed: OSD and subtitles went black when using error diffusion
* fixed: subtitle depth was always set to 4, even for 2D subtitles
madshi is offline   Reply With Quote
Old 22nd February 2016, 11:24   #36380  |  Link
ryrynz
Registered User
 
ryrynz's Avatar
 
Join Date: Mar 2009
Posts: 3,481
Quote:
Originally Posted by madshi View Post
I wondered the same thing. Upscaling AR filter was not modified in any way.
Think I know what happened there. Thanks, I checked it and it's as you say.

Quote:
Originally Posted by har3inger View Post
Wait, was the AR filter for upscaling updated as well? Changelog just says it's for downscaling, but the results you show here are clearly different. Is this jinc downscaling?
Nope, image editor quirk. Should play it safe when it comes to layers.
ryrynz 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 06:39.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, vBulletin Solutions Inc.