Un-biased quality comparisons.

[PuzZLeR]

I'd like to compare quality of my H.264/AVC encodes between different implementations (x264/Recode/etc.) or different methods (filters, CABAC, etc.) with an UN-BIASED decoder/player/decoder.

VideoLAN seems to have better quality for x264 encodes.

Nero ShowTime seems to have better quality for Nero Digital encodes. It only accepts certain files as well.

QuickTime? Let's not go there. I could write pages on this one.

With DivX/Xvid, it's easy. Just fire two instances of VideoLAN and time them and do your comparisons visually.

With AVC it's not the same. VideoLAN is biased. But as well, VideoLAN seems to smooth out other playback windows when more than one is running on my PC. Maybe it's the higher complexity of AVC that it can't handle in multiple fashion. Not sure.

Is there a method, preferably with two windows running, that can play the clips side-to-side to compare?

Thanks in advance. Very helpful forum so far.

[akupenguin]

wtf?
H.264 is the one that's easy to compare; all H.264 compliant decoders (that support the features in your stream) produce identical output. Use the JM reference codec if you're you're really worried, but first I ask to see your proof that there is even a difference (let alone bias) between decoders.
MPEG4ASP is the one that's hard to compare fairly, since Xvid and Divx use different idct approximations.

[bond]

given that postprocessing is disabled every decoder/player should output exactly the same quality for a specific stream

[PuzZLeR]

My apologies!

My PC was the problem. When running multiple instances of AVC playback, it somehow played the first video fine, but blurred out the second one.

Anyhow, I’ve re-arranged the orders around (clips and players) and continued to get the same problem, even with MPC as well.

I thought it was a bias at first, especially since this screw-up didn't affect my ASP playbacks but when I noticed ffdshow and WMP playback got all weird with AVC I figured it out.

Anyhow, the problem is that yesterday I installed a trial version of a “certain software” for H.264 and it messed up my PC with AVC, so when I attempted testing I got weird results and I thought it was some bias. Thank goodness I use Restore Point before every installation. Everything is normal now.

(I won’t mention what I installed since it’s a reputable firm and I don’t want to break any forum rules or mess their rep because I may have screwed up somewhere.)

Nevertheless, does anyone know a good way to test comparitive playback of two or more encodes, particularly in comparing certain filters, etc. that is simple and straightforward instead of firing up multiple instances of players?

I downloaded the JM software as akupenguin mentioned, but I honestly have no clue what it does or how to run it or anything. And I don’t understand the MSU Quality Measurement Tool either. Huh?

Thanks for your patience with my learning pains folks. It’s very much appreciated.

[delacroixp]

Nevertheless, does anyone know a good way to test comparitive playback of two or more encodes, particularly in comparing certain filters, etc. that is simple and straightforward instead of firing up multiple instances of players?

My favourite method is still to get a frame snapshot from each encode, place them in a folder and run a slide show untill you spot the difference.
The problem is that you need the exact same frame-number and most players don't have that feature.
Also, individual frames are very static and miss the overall "movement" of film.

I use Zoom Player (http://www.inmatrix.com/) but unless you have a pretty large screen (or two) it's difficult to place them side by side.
A slow computer (poor GPU) will also have problems running 2 movies simultaneously... especially at full res.


:):devil::D
Pascal