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. |
14th August 2012, 05:28 | #13701 | Link |
Broadband Junkie
Join Date: Oct 2005
Posts: 1,859
|
Lanczos3 anti-ringing for Luma does seem like a good option with madVRantiringing6. Still a bit softer than normal Spline36 on the 720p sample because of the issue ryrynz mentioned, but it retains fine contrast slightly better than Spline64 anti-ringing. Though both Spline36 and Spline64 seem considerably better at interpolation of gradients, at the expense of less fine contrast.
As madshi mentioned, both Lanczos4 & Lanczos8 do show residual dark/bright ringing with the test6 build. @madshi Would it be possible to add some sharper Spline interpolators to madVR for testing with anti-ringing? Spline144 (6-tap 12x12) and Spline256 (8-tap 16x16)? Related thread about calculating Spline coefficients & avisynth filter source code |
14th August 2012, 19:08 | #13704 | Link | |
Registered User
Join Date: Mar 2002
Posts: 2,323
|
Quote:
Tested on a 1080i@30fps clip too ...
__________________
Ryzen 5 2600,Asus Prime b450-Plus,16GB,MSI GTX 1060 Gaming X 6GB(v398.18),Win10 LTSC 1809,MPC-BEx64+LAV+MadVR,Yamaha RX-A870,LG OLED77G2(2160p@23/24/25/29/30/50/59/60Hz) | madvr config |
|
14th August 2012, 21:17 | #13705 | Link | |
Registered User
Join Date: Dec 2005
Posts: 32
|
Quote:
Sorry for bumping |
|
15th August 2012, 00:31 | #13706 | Link |
Registered User
Join Date: Mar 2009
Posts: 3,650
|
Madshi, if you're having to compromise anti-ringing quality for performance could you allow for a HQ option? I'm only interested in overall improving quality, I don't care about the performance hit, I'll upgrade if I need to.
|
15th August 2012, 05:03 | #13708 | Link |
Registered User
Join Date: Mar 2011
Posts: 60
|
Aye.
My preference is to have minimum amounts (although as accurate to the source as possible as well) of aliasing with modest (although as minimum as possible, if possible) amounts of artifacts (ringing, blockiness, etc.) while maintaining maximum sharpness and detail. The current anti-ringing (build 6) seems to be very good, while using LZ3 with it on for both chroma and luma (with my preferences, which seems to be similar to many others). But if it can be done better (in whatever way) with some performance loss, even if it's only a little better IQ, it'd be worth it to have some kind of "HQ" option Last edited by Hprd; 15th August 2012 at 05:06. |
15th August 2012, 16:41 | #13709 | Link |
Broadband Junkie
Join Date: Oct 2005
Posts: 1,859
|
@ryrynz
I took a look at your sample, and it seems like the source frame does have faint areas on the lines. madVR anti-ringing appears to be preventing the dark ringing which actually had a positive effect on the lines of that sample, while the anti-ringing result may actually be a closer representation of the source pixels. I wonder if this is actually solvable without negative ringing side-effects on sharp high-contrast edges. 704x480 -> 1706x960 (800% crop): Nearest Neighbor vs Spline36 Nearest Neighbor vs Spline36 anti-ringing Spline36 vs Spline36 anti-ringing |
15th August 2012, 23:04 | #13710 | Link |
Registered User
Join Date: Mar 2009
Posts: 3,650
|
I think what we can take away from this is that when resizing the most accurate isn't always the best solution. It thus looks like a good sample to use for refining anti-ringing the result of which could address some of your concerns also. I hope something good can come from it, eagerly awaiting your progress Madshi.
|
16th August 2012, 20:51 | #13711 | Link | |
Registered User
Join Date: Aug 2012
Posts: 32
|
Quote:
- Bitstream is possible with ffdshow and other software. - 3D support is possible in SBS... in MPC-HC. It's the MVC codec who isn't supported yet in MPC-HC. |
|
17th August 2012, 03:07 | #13712 | Link |
Registered User
Join Date: Jan 2009
Posts: 1,210
|
What would 10bit actually provide or look like if not dithered down to 8bit? I know the benefits of encoding to 10bit (less banding, better compression, etc.) but since it gets dithered down to 8bit for the majority of displays which happen to be 8bit, I'm unsure about what I'm missing by not having a 10bit display.
|
17th August 2012, 03:11 | #13713 | Link | |
契約者
Join Date: Jun 2008
Posts: 1,576
|
Quote:
with real 10bit source (not from bluray or dvd) you maybe will see better picture if your eyes can distinguish more than 17 million colors though i'm not aware of such sources other than some artificially created stuff where 10 bits matters mostly during processing. or actually i use up to 32bitdepth sometimes. not that it matters after stuff was rendered... Edit: btw afaik there is no software that will allow you not to dither. whatever dither or truncate. Last edited by Keiyakusha; 17th August 2012 at 03:44. |
|
18th August 2012, 02:31 | #13714 | Link | ||||
Registered User
Join Date: Mar 2009
Posts: 962
|
Quote:
Quote:
Quote:
Quote:
__________________
MSI MAG X570 TOMAHAWK WIFI, Ryzen 5900x, RTX 3070, Win 10-64. Pioneer VSX-LX503, LG OLED65C9 |
||||
18th August 2012, 02:53 | #13715 | Link |
契約者
Join Date: Jun 2008
Posts: 1,576
|
You mean if TV somehow accepts 10bit input and dithers it (or whatever these TVs do, cause they may do a lot) will bring some benefits over proper dithering like the one madvr does? How come, the only reason x264 creates 10bit if configured to do so is to put some noise there that it will otherwise throw away. leaving dealing with that to some quality improving (damaging?) algorithms of modern tvs doesn't sounds too good.
Last edited by Keiyakusha; 18th August 2012 at 03:00. |
18th August 2012, 02:57 | #13716 | Link |
Registered User
Join Date: Jan 2012
Posts: 106
|
Hi,
I am using potplayer, lav codecs and reclock audio and madvr video renderer. In madvr I use windowed mode (exclusive is jumping in and out when I go to the menu or if I want to use the bottom bar which is annoying) and from what I have read I should disable desktop composition, to avoid vsync problems, but that actually results in picture tearing (near the bottom of the screen). If I don't disable "aero" there is no tearing but I have a few dropped frames at the beginning of the playback and occasionally if I pause and restart the movie. Advice appreciated. edit: I tried in MPC-HC as well and it behaves the same. |
18th August 2012, 03:03 | #13717 | Link |
契約者
Join Date: Jun 2008
Posts: 1,576
|
Maybe I'm wrong but I always though this is how it is and how it should work. Why would you care about drops if you pause anyway? And no surprise in aero actually helping. dunno why, but this is the case for many users (not only madvr). It have its own vsync
Last edited by Keiyakusha; 18th August 2012 at 03:05. |
18th August 2012, 10:25 | #13718 | Link | |
Registered User
Join Date: Mar 2009
Posts: 962
|
Quote:
__________________
MSI MAG X570 TOMAHAWK WIFI, Ryzen 5900x, RTX 3070, Win 10-64. Pioneer VSX-LX503, LG OLED65C9 |
|
18th August 2012, 12:25 | #13719 | Link | |
Registered User
Join Date: Jun 2006
Posts: 452
|
Quote:
With a 10 bit yuv420 (same for 422 and 444) the banding would be much less and probably not noticeable any longer. If you keep the whole chain from source -> display on 10 bit, you would have a more accurate image I guess, with no banding. This would require the monitor and video card to accept and output 10 bit (of course not 10bit RGB, asthis is nonsense) Last edited by Pat357; 18th August 2012 at 12:42. |
|
Tags |
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling |
|
|