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. |
22nd February 2014, 14:09 | #23663 | Link |
Registered User
Join Date: Sep 2006
Posts: 2,197
|
I dont really have a clue about all those test patterns and stuff, but Id suggest the following anyway:
- since madVR is used as a video renderer, comparisons should almost exclusively be made with videos and not some kind of grey ramps. its rather pointless to use something else as comparison for which madVR is not meant to be used. - it would also help if madshi could state each time what to look out for when comparing stuff, which factors and which others can or should be disregarded. e.g. in case of error diffusion low noise smoothness is what should be achieved and not crispness. so the factor crispness should be disregarded when doing comparisons no matter how good it might subjectively look or some kind of pop effect or whatever. imho its vital that people stick to this when doing their comparions, but also that madshi points out such things each time this should apply. otherwise all this testing wont lead to anything.
__________________
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) |
22nd February 2014, 15:08 | #23664 | Link |
Registered User
Join Date: Aug 2007
Location: Fort Wayn,Indiana
Posts: 52
|
madVR - high quality video renderer (GPU assisted)
OCD for me as a4 has more pop but OCD is more detailed and accurate and a deeper and more detailed depth going back into
the screen. More natural or balanced. All are good and if we have access to all the test builds In one folder we can pick and choose. Problem solved for everyone with only one build needed. Thanks everyone as we all have different eyes and viewing. We need to all use the same setting as madshi if he would share so we can give input that is similar. I know screens and video cards are different but if we use default for MadVR and our graphic cards would that not level the playing field? We need a standardize test as not everyone has the same gpu power. Since I am not a programmer I am just going to leave my input as OCD or R2D2 and a brief description. You all know that I prefer the GTX770 over the 7870xt on my projector and that is with everything default on both madVR and gpu settings. Where the GTX770 pulls ahead is when gpu load increases and remains smooth and stutter free with no over heating with neurons used. Sorry for the additional comments I like testing but sometimes feel it is for my own selfish needs.
__________________
2011 VDC 9500LC ULTRA Mike Parker Modded 12-3-2015 GTX-770 and Intel Xeon V1235 Windows 10 Last edited by GREG1292; 22nd February 2014 at 16:25. Reason: spelling and grammer issues. |
22nd February 2014, 15:19 | #23665 | Link | |
Registered User
Join Date: Sep 2013
Posts: 919
|
Quote:
With no color at all (Saturation to -100). I can see colored pixels the MultiColor dithering creates in 4-bit with real content. Then I quickly changed the desktop folder name to something else to return to 8-bit (no need to close the paused video). Now I can't see any color what so ever. To verify that the colored dithering is still there in 8-bit (although completely unseen now), I took some screen shots of 4-bit & 8-bit (same frame) and boosted the Saturation in photoshop. Yes, the 8-bit has the same colored dithering but I needed to boost the saturation four times more compared to 4-bit, to see the same level of saturation. Still, OppositeColor Dynamic is absolutely the smoothest looking, in 4-bit or not. P.S On 4:2:2 TV's the colored dithering pixel will be TWICE as big horizontally, that may bias some users to falsely judge the Multi/Opposite Color builds. So please try to test the builds on a 4:4:4 monitor or TV.
__________________
System: i7 3770K, GTX660, Win7 64bit, Panasonic ST60, Dell U2410. Last edited by James Freeman; 22nd February 2014 at 15:43. |
|
22nd February 2014, 16:20 | #23666 | Link | |
Registered User
Join Date: Dec 2013
Posts: 753
|
The only convincing alternative I can think of is the one I posted a few pages back where you simply pick a random color (r,g,b) such that r+g+b=0. The Yuma noise from this algorithm should be significantly lower than the one from the opposite algorithm, all channels have the same distribution, and all values between -0.5 and 0.5 are possible. The only downside I've found is that not all values between -0.5 and 0.5 are equally likely, values around 0 are twice as likely as those at the extremes (this is because there are twice as many possible values for g,b such that r+g+b=0 if r is near 0). This effect may actually not be a bad thing but you'd need to check that gradients still look nice and possibly retune the algorithm to account for the different noise distribution.
Quote:
Edit: I actually made a mistake when measuring the amount of Y noise, the oppositeColor method seems to be better than the alternative I proposed. I thought it was weird that there didn't seem to be much of a difference between oppositeColor and multiColor was so small, turns out that the numbers I used for oppositeColor weren't random. Anyway I now have a mathematical 'argument' that oppositeColor causes less Y noise than multiColor. Last edited by Shiandow; 22nd February 2014 at 20:36. |
|
22nd February 2014, 17:03 | #23667 | Link | |||||||||
Kid for Today
Join Date: Aug 2004
Posts: 3,477
|
I might be mistaking but some of you would appear to base their opinion on the 4bit trick? Nev has made quite clear that this doesn't quite scale with 8bit...maybe madshi could provide a way to go 6bit? It would be closer to 8bit and might actually be useful to the 6bit TN crowd(FWIR mVR doesn't currently officially support ED for 6bit).
Quote:
Also, several Sammy 1080p TV reviews claim that if you disable BFI you only get 360p.. Quote:
Quote:
Quote:
Quote:
Quote:
Quote:
If it's a deal breaker or if all you care for is pop "über alles" then A4 is a good choice too IME/HO. Hopefully at some point madshi with decide to go with a default algorithm and allow the minorities to pick their favorite using a registry key...if providing that choice in the UI is truly out of the question. Quote:
So does that means that the softer look & motion blur of the color builds is by design and more "technically optimal"? All this said, I come from the CRT/DLP world and I basically want my LCD to look like them. Many ppl claim that DLP comes with a stunning "cinema-like" analog look and I was strongly impressed by my old 3K:1 Mitsubishi HC3100 on a daily basis using a 2m wide projection screen from a 3.3m distance. It looked so real and so enjoyable. PQ was a real thrill thanks to the 600:1 ANSI contrast and invisible black during bright scenes(with its iris closed). DLP looks very sharp thanks to the single panel(and mineral glass lenses on the HC3100) but also grainy due to the mirrors holders and it's a known trick to use grain in order to deblock. Also, FWIR even the most optimal bluray 8bit encodes are always a compromise when it comes to movie grain...all this to say that subjectively speaking A4 does it all at once for me: deblocking, natural looking grain-based edge enhancement, analog look, reinforcing movie grain without changing its pattern/direction, etc etc.. A4 also saves the day with upscales and I still believe that every display comes with its own dithering pattern and it's obvious that we all more or less see differently so there can't be such a thing as a magical "one size fits all" mVR dithering algorithm IMHO. Also, even with BFI LCD is still not nearly as responsive as CRT. Quote:
I find 32" from a 80 cm distance to be perfect for 1080p content, but it'll sure take something seriously big to start whining all over again about 4K dithering to madshi Thank you for the kind words, I learned a lot from ppl like MLill on HCFR. Last edited by leeperry; 22nd February 2014 at 18:17. |
|||||||||
22nd February 2014, 17:55 | #23669 | Link |
Registered User
Join Date: Apr 2012
Posts: 169
|
Does anything (besides the porting effort) speaks against switching to openGL+openCL and go multiplatform? I think you mentioned going pure openCL at some point.
__________________
AMD Phenom II X4 970BE | 12GB DDR3 | nVidia 260GTX | Arch Linux / Windows 10 x64 Pro (w/ calling home shut up) |
22nd February 2014, 17:57 | #23670 | Link |
Kid for Today
Join Date: Aug 2004
Posts: 3,477
|
Well, +500:1 ANSI for a video projector is quite high AFAIK...at least it was when I was using it in 2007. And I wasn't dressed like a ninja in a room with black walls ^^
MLill kept a measured specs comparison here and you can see that even in 2012 there are few projectors that would provide >500:1 measured ANSI contrast, most of them being DLP for that matter. I of course thoroughly enjoy how on a LCD flat screen the ANSI contrast is equal to the native contrast |
22nd February 2014, 18:03 | #23671 | Link | |
Registered Developer
Join Date: Mar 2010
Location: Hamburg/Germany
Posts: 10,344
|
Quote:
So a feature loss, for what gain? Its not like other platforms have DirectShow, or even any comparable video frameworks where a renderer could just "plug in" like it can in DirectShow. Not to mention that Error Diffusion uses DirectCompute now, which is part of DirectX, and a Windows technology.
__________________
LAV Filters - open source ffmpeg based media splitter and decoders |
|
22nd February 2014, 18:07 | #23672 | Link |
Registered User
Join Date: Mar 2007
Posts: 934
|
I've never understood the love for projectors. Sure it's a relatively cheap way to get huge pictures but it's impossible to get as good black levels as most LCDs, let alone plasmas, even in dark rooms. I remember watching clips of LotR at a friend's house a few years ago on his projector and I wasn't impressed at all.
__________________
TV Setup: LG OLED55B7V; Onkyo TX-NR515; ODroid N2+; CoreElec 9.2.7 |
22nd February 2014, 18:17 | #23673 | Link |
Registered User
Join Date: Dec 2008
Posts: 18
|
currently comparing the image output of the various experimental builds- but have to echo other people's views in that its great to even be able to try out different flavours of picture processing - especially as some is personal taste really
with the new processing, OpenCL upscaling etc - I've not seen a better image output to my Panasonic Plasma (Pro PF11) - and therefore would like to express my great thanks to Madshi |
22nd February 2014, 18:30 | #23674 | Link | ||||
Registered User
Join Date: Apr 2009
Posts: 1,019
|
Quote:
But if you're trying to critically judge something like dither algorithms where pixel-level details matter, I wouldn't do it in anything less than full resolution. Quote:
Every time I think I am seeing noise in the image from dithering, it turns out to be noise that is encoded in the source. (just as visible when I disable dithering) In fact, I am finding that due to the noise level being lower with error diffusion rather than random dithering, that sort of thing is a lot more noticeable now. Quote:
Quote:
|
||||
22nd February 2014, 18:37 | #23675 | Link | |
Registered User
Join Date: Dec 2008
Posts: 496
|
Quote:
However, the images that were provided (e.x. by you and 6233638) clearly show that more noise is added, not less. Thereīs more noise dots and they also are closer together, meaning, itīs way thicker than with the ED4/mono_static build (thatīs the smoothness youīre talking about), so basically, itīs filling way more gaps as before with the ED4/mono_static build. And since we are only comparing still shots (where this is already visible), when being in motion, thereīs even more noise which is changing itīs pixel position all the time, because the noise is dynamically changing, whereas it was static before. That doesnīt have to be a bad thing and I didnīt say as much. Actually I already said the opposite myself, before I even did a subjective test, because when I am thinking about changing picture contents, the noise should adapt to the contents or else, we wouldnīt technically be dithering anymore, we would just add noise. The problem with this is, that on static parts of an image (where there are usually quite a lot of), you would have a lot of changing noise all the time. This was the trade-off we were talking about. Now, concerning the noise itself. If itīs visible or not, is a completely different story. And thatīs all I was trying to say. Now, the question is, what we want to achieve and I agree with you, we should strive for an accurate appearance. But how do we know that we are accurate? Shouldnīt the dithering algorithm only add noise where it is needed? Why donīt we add even more noise that is closer together? And I was just thinking out loud, if we already crossed that bridge, yet. Now, comparing the builds, while the multicolor_dynamic build also makes that additional noise stand out quite a bit (thatīs a bad thing), the opposite_dynamic build reduces the visibility of the noise by quite a lot. So, yes, the opposite build clearly has a lower perceived noise floor compared to the multi_dynamic build. But ED4 still has the least amount of noise (just look at the sky picture, provided by 6233638), even though itīs clearly more visible. Which is the trade-off with ED4. Since we are currently evaluating the medium-noise dithering algorithm (which the opposite_dynamic build comes closest currently), the aim should be "as accurate as possible, like RD with a lot less noise", while the aim for the low-noise algorithm should be "accurate, but only as much noise as needed". Or how would you define that yourself? Last edited by iSunrise; 22nd February 2014 at 19:14. |
|
22nd February 2014, 18:45 | #23676 | Link |
Registered User
Join Date: Dec 2008
Posts: 18
|
for a Panasonic Plasma (PF11 with DualHDMI board) - should I be aiming to output 4:4:4 from my ATI card or 4:2:2 ? I thought my panel accepts RGB 4:4:4 - but I struggle to get to see any of the black levels in the black levels test
|
22nd February 2014, 18:57 | #23677 | Link | ||||
Kid for Today
Join Date: Aug 2004
Posts: 3,477
|
Quote:
Quote:
All those static screenshots are undeniable, especially as madshi made clear that the goal is softness.....too bad those colorful builds look blurry in motion to me(and to a few others as well apparently). Quote:
Quote:
Last edited by leeperry; 22nd February 2014 at 21:02. |
||||
22nd February 2014, 20:01 | #23678 | Link |
Guest
Posts: n/a
|
OppositeColor looks much cleaner than others to me. I don't know why my opinions differ so much from others, maybe its the 4:2:2 3000:1 CR SPVA that makes things worse with noisy builds or whatever else, but I like OppositeColor the opposite the most... I also like the least noisy ED build the most (Noisy2Limited I think is the name).
Should we not have a lot more people in this sample group to really figure out what the public likes the most? |
22nd February 2014, 20:25 | #23679 | Link |
Guest
Posts: n/a
|
Also, I wonder if the broken OpenCL <-> D3D9 interlop is because nVidia does not feel like supporting D3D9 is very important. I think Windows XP is too ancient to support. Who is going to use ED and NNEDI3 on Windows XP??? You need a powerful card to run it and all powerful cards users have Windows 7 or Windows 8.
|
22nd February 2014, 20:29 | #23680 | Link | |
Kid for Today
Join Date: Aug 2004
Posts: 3,477
|
Actually I don't think you can get OCL support under XP at all, at least it wasn't recognized as active by GPU-Z with a HD7850 and the latest AMD drivers a few months ago: https://bitcointalk.org/index.php?topic=117956.0
Quote:
|
|
Tags |
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling |
Thread Tools | Search this Thread |
Display Modes | |
|
|