AVS Forum banner
Status
Not open for further replies.
1 - 20 of 23 Posts

·
Registered
Joined
·
265 Posts
Discussion Starter · #1 ·
Here u get some hard facts about dvd decoder quality, just compare/download the pictures and see for yourself.


System setup:

Windows XP + dx9b (proppage.dll registered (from dx9 SDK) for dx9b video render settings and screenshot ability in vmr7/9) nvidia 53.03 WHQL driver (no special settings)

Zoomplayer 3.30 (costum dvd mode)


GFX: GeForce FX5700 ultra (DX9 card)

DVD: Alien special edition, PAL version, THX optimizer included

Decoder versions
  • Elecard : mpgdec.ax, 2.0.0.2510
  • WinDVD : IVIVIDEO.ax, 5.3.5.22
  • Sonic : CinemasterVideo.dll, 2.5.03.0808 (TheaterTek use the same decoders)
  • Cyberlink: CLVSD.ax, 5.0.1031


DXVA was disabled in all decoders, so its plain software mode.


Zoomplayer setup

Zoomplay 3.30 dvd costum mode playback render set to VMR7 aka Video Renderer|{6BC1CFFA-8FC1-4261-AC22-CFB4CC38DB50}

Why? cause u can take screenshots direct out of the render prop. page in dshow. So u get a perfect screenshot without using crappy programms.

I did not use VMR9 since its still bugged and i wanted do the test in native yuy2 mode. But vmr9 + nvidia + yuv modes = buggy. The vmr7 render just work fine like the overlay mixer. VMR7 is also the standard render windows xp dshow set's!


THX testpattern 01, deinterlace test


[edit]I removed the first testpattern since it had nothing to do with deinterlace power, it was just a bad pattern since the deinterlace routine thinks the image is interlaced but it's not.[/edit]


THX testpattern 02

http://img17.photobucket.com/albums/...222/THX_02.jpg
http://img17.photobucket.com/albums/...22/THX_02c.jpg



300% zoom

http://img17.photobucket.com/albums/...22/THX_02a.png



Result: In that test all decoders show that they can produce correct colors and linear lines, the overall quality of all 4 are good but u can spot differences in the zoomed images. Its hard to tell what image looks best, all have cons and pros if u compare the zoomed shots.


[Edit]added cyberlink results and looks oki, what u cant see is that cyberlink looks some kind of crappy in some areas if the movie plays (some blocking artefacts)[/Edit]
 

·
Registered
Joined
·
265 Posts
Discussion Starter · #2 ·
Part2:

bright Movie scene 03

the aspect ratio is wrong cause i could not force the VMR7 render to output the screenshot with a different AR than it gets from the dvd, its np since the image quality will not change with diff. aspect ratio


http://img16.photobucket.com/albums/...Elecard_03.jpg


http://img16.photobucket.com/albums/.../WinDVD_03.jpg


http://img16.photobucket.com/albums/...2/Sonic_03.jpg


http://img16.photobucket.com/albums/...berlink_03.jpg


results:

If u download the .png pictures and compare them with a tool or with the XP build in picture preview'er u will see that u cant realy spot any differences at all. U can just spot some very very small differences at the lower body skin regions, if u have good eye's.

U may say: "hey that are all 3 the same image u did not change the decoder in zoomplayer." trust me i did change them.

So i compared all 3 org. .bmp screenshots bit by bit to see if there are any diff. at all.


[Edit] added cyberlink shot and we can easy see the image is unsharp if u compare the face, i tested also the org. player and overlay mixer/vmr9 mode all produce the same unsharp image[/Edit]


Elecard vs WinDVD = 4279 diff. bits

Elecard vs Sonic = 4600 diff. bits

WinDVD vs Sonic = 7972 diff. bits

Cyberlink vs Elecard = 982 diff. bits

Cyberlink vs Sonic = 1759 diff. bits

Cyberlink vs WinDVD = 2829 diff. bits


[Edit]That average offset was crap. Sorry it has nothign to say about image quality, since some bits may differ more than others. The comparsion just show that even if we cant spot much diff. there are some in the bit matrix[/Edit]


So we dont have a winner in that scene, Elecard/WinDVD/Sonic all perform the same image quality, only Cyblerlink produce a unsharp image.


detail Movie scene 04


http://img16.photobucket.com/albums/...y22/Ele_04.jpg


http://img16.photobucket.com/albums/.../WinDVD_04.jpg


http://img16.photobucket.com/albums/...2/Sonic_04.jpg

http://img16.photobucket.com/albums/...berlink_04.jpg


results: Again if u can spot any diff. at all u must have phat eye's. The images look all the same if u compare them. U can spot some minor diff. at the upper right side in the backround with the WinDVD decoder.


[Edit] added cyberlink image, and like in the scene before the image is unsharp[/Edit]


The bit by bit comparsion showed this:


Elecard vs WinDVD = 17901 diff. bits

Elecard vs Sonic = 13110 diff. bits

WinDVD vs Sonic = 16908 diff. bits


Seem in that image we have more diff. beetwen the decoders if u compare them bit by bit.


But again we dont have a winner all images look the same, except for teh unsharp Cyberlink image.


So what this all means and what is with ffdshow rezise and filters?


[Edit] i removed my personal results, everyone can compare the image themself [/Edit]


So why i did not test with ffdshow reziser and other filters u normaly use? Easy since ffdshow is just added AFTER the org. mpeg decoder filter, it just matters what ffdshow get's handed over to work with. So i just wanted to see what ffdshow gets as input by the decoders. Means if the decoder produce a crappy image ffdshow will also perform worser.



[Edit]i will run a test with ffdshow rezise and diff. renderes to see if the cyberlink results can be compensated with more ffdshow sharpness[/Edit]


May this test help u to choose your decoder in zoomplayer, i did not take into account what features the org. player software has. If u dont use zoomplayer all 4 products may differ more!
 

·
Registered
Joined
·
23,131 Posts
Very interesting, I assume you ran all these in SW not DXVA mode, correct. A comparison of SW vs DXVA would also be interesting. It would be kind of nice to see some shots from nVDVD as well, that decoder has a lot of fans.


Good job, thanks, should provoke some interesting conversation.
 

·
Registered
Joined
·
3,137 Posts
Thanks a lot for taking the time to do the test. I've tried different players,

and have settled with PowerDVD 5 because it can play many formats. It

also make overlay setting easier.


Nice,


Nicholas
 

·
Registered
Joined
·
265 Posts
Discussion Starter · #7 ·
Quote:
Originally posted by MagnusB
Why not add WinDVD with overlay and ffdshow resize?

And Sonic with WMR9 just to show the differences.


///Magnus
What that will proofe than? It will proofe that rezise+overlay is better than plain VMR9 thats all? I think we all know that already.

Maybe i will do a overlay vs vmr7 vs vmr9 test later. Still have to find a decent way to correct the color mixer diff. in each render and also that crappy nvidia vmr9 bug.


@stanger89 yes i tested it with software mode, testing dxva might be a waste of time since nearly all use ffdshow and it has problems connecting to the decoder if dxva is enabled and i never will play anything back again without ffdshow :) i just love level fix and dering postprocessing too much.

Adding nvdvd decoder is a waste of time atm since the new 3.0 version will be ready soon, but im not a beta tester so i dont have that decoder/player.


@intinig if u just use the decoder in zoomplayer than it wase a waste of money sorry :( if u realy use windvd player u also got a better sound decoder + maybe that audio booster pack wich adds tonns of cools stuff. For pure zoomplayer quality u see the results sorry.
 

·
Registered
Joined
·
2,548 Posts
I think this comparison is impractical...


What you should do is take an uncompressed video source, encode it using the best encoder (That would be CCE I believe) then decode it and compare the results ot the original uncompressed image. That way you'll have a reference to test against.


BTW: ZP does have internal screenshot support when using VMR9 (Alt+F). Works well for DVDs.
 

·
Registered
Joined
·
590 Posts
Quote:
Originally posted by AndyIEG
First that means that all 3 decoder will result in the same "static" image quality, we just compared static screenshots but for me also the playing movie "feeled" the same.

.

.

.

Final words: Since the images are so damm identicaly to eachother i must admit that its a myth that WinDVD decoder is the "best" or that "Sonic" aka TheatherTek owns all other. It all ONLY depends on the ffdshow/player/overlay color settings!

It also showed that the Elecard decoder produce the same quality as WinDVD/Sonic. U also have take into account that the elecard has a nice build in postprocessor filter and u dont need any registry tricks at all to get it work with ffdshow.
Sorry, but your testing miss the points... :rolleyes:


Here are the most important differences between the three:


MPEG Blocking

Blocking comes due to less accuracy when decoding. The decoder produces errors which accumulate between the I-Frames and leads to blocking artefacts. Only more bits can better the problem.

Elecard: best of the three, produces nearly as less blocking as a good standalone

WinDVD: not much blocking, but still a bit more than Elecard

Sonic: horrible Blocking against the other two! Only if you use DXVA on a new Radeon (9500 and up) blocking is on Elecard level.


Chroma Upsampling

Elecard: does simple line doubling when converting 4:2:0 to 4:2:2 and outputting via YUY2. This leads to horrible chroma aliases (stair steps). Clearly visible with e.g. "Monster Inc" oder "Toy Story". The 200% zoom color test pattern in this thread also shows this (look at the horizontal edges).

Only if you connect it to ffdshow and go with YV12 to the overlay mixer the chroma upsampling is done in hardware and looks perfectly.

WinDVD: has the chroma upsampling error when outputting YUY2. Only if it connects with YV12 to ffdshow the chroma channel is right.

Sonic: does chroma aliasing the same way as Elecard.


When using DXVA WinDVD and Sonic show chroma the right way and also show extremely little blocking artefacts (on newer Radeons only). When using an old Radeon the blocking is as much as with Sonic in software mode!


Banding

I have to research this a bit more. I just know that the older versions of WinDVD (4 and less) showed horrible banding. Sonic and DXVA are OK.
 

·
Registered
Joined
·
52 Posts
@ FoLLgoTT


Basically from your comments the best decoder is sonic in dxva on a radeon 9500+. How do you fit ffdshow in the equation? Is it possible to use ffdshow to avoid these decoder bugs?
 

·
Registered
Joined
·
265 Posts
Discussion Starter · #12 ·
Update: added cyberlink results, u have to flush your offline browser cache to see the new images.


@Blight take into account that this was no test against the org. uncompressed video, we normaly dont watch uncompressed dvd. This test just shows screenshots of the same frames decoded with diff. decoders. Not more not less. I dont think comparing the decoders against the org. source will help us choose our decoder. We have to choose from what we can get and see. The comparsion is beetwen the 4 decoders thats all.


@FoLLgoTT thx for the input, i know about that chroma bug but dont have any dvd with that bug to test ... sorry also that bug isnt visible anytime and on every dvd.


For the MPEG Blocking u mean the ordenary blocking wich can be compensated with good deblock postprocessor's right?
 

·
Registered
Joined
·
590 Posts
@intinig

When using ffdshow DXVA cannot be used. So you have to choose the decoder with the best software decoding. Until know this is Elecard.

But there occurs one big problem when you let ffdshow do the resizing. The YV12 trick doesn't fit, because the chroma aliases produced by Elecard are also resized! This is not a too big problem, because the Lanczos smoothes the aliases a bit, but they are still there.


Also only the newest version of the Elecard supports subtitles. But again it has the chroma bug.


I think the decoder after Elecard is WinDVD 5 in software mode. But I have to test it a bit more.


Btw, NVDVD 2.55 looks terrible! I was really surprised, it is the only decoder which also shows MUCH blocking with DXVA enabled! Mosquito noise and banding are also absolutely inaccaptable. I hope NVDVD 3 will be much better.
 

·
Registered
Joined
·
590 Posts
Quote:
Originally posted by AndyIEG
For the MPEG Blocking u mean the ordenary blocking wich can be compensated with good deblock postprocessor's right?
Yes, I mean this sort of blocking. But to do a post process is not the right way, because it eliminates also details. The right way is to implement a more accurate decoding engine. In this point the decoders differ much. And as I stated also the decoding with Radeons (new/old core) differs much.
 

·
Registered
Joined
·
265 Posts
Discussion Starter · #15 ·
Quote:
Originally posted by FoLLgoTT
Yes, I mean this sort of blocking. But to do a post process is not the right way, because it eliminates also details. The right way is to implement a more accurate decoding engine. In this point the decoders differ much. And as I stated also the decoding with Radeons (new/old core) differs much.
ah oki i have much experience with deblock/dering postprocessors watching xvid/divx ripps. If u setup the deblock postprocessor the right way there wont be visible loose of details. The problem is if u use ffdshow postprocessor for deblock u just have to select nic's and set x to 30. If u also enable mplayer deblock or Nic's dering u will loose details. So for xvid i can use the deblock the decoder filter itsself has build in and than just use the mplayer dering with nic's disabled in ffdshow.


Im not sure if the better deblock resluts from Elecard might be cause u had the build in postprocessor enabled?
 

·
Registered
Joined
·
590 Posts
Quote:
Originally posted by AndyIEG
Im not sure if the better deblock resluts from Elecard might be cause u had the build in postprocessor enabled?
But I'm sure. ;)

The post processor is always disabled here. When I enable it I loose a small amount of detail. Blocking does not get much less, because there is nearly none also without post process.


I made these Screenshots a time ago and posted it in a german homecinema forum.

WinDVD was version 4 and shows much banding. Sonic with DXVA runs on a old Radeon and shows much blocking. The videolan shot looks like Elecard, because both use the same libmpeg2 decoder. :)
 

·
Registered
Joined
·
265 Posts
Discussion Starter · #17 ·
ah nice screenshots, i realy need to find a better scene to compare blocking better :)


PS: atm im trying to verify a color conversion bug by using ffdshow reziser... not sure if its the nvidia bug again...
 
1 - 20 of 23 Posts
Status
Not open for further replies.
Top