Lower quality with VC-1 source than AVC

[buffyangel108]

Currently backing up my blu-ray collection (mixture of VC-1 and AVC .m2ts files) to 720p x264 mkv to save space on my media server.

I've noticed that whenever the source is VC-1, even if that source is pristine, the resulting x264 stream always contains a fair amount of added noise, even if the bitrate and profile should be more than enough to give near-transparency at 720p (~15000kbps, veryslow/placebo). This noise isn't introduced by resizing, as it's also present if I do a 1080p encode. AVC source files are unaffected too: it's only VC-1.

Along similar lines, if I choose to encode in CRF mode, I've noticed that if the source is VC-1, I have to use a lower CRF (e.g. 14) to achieve the same quality (and average filesize) as CRF 16 would give with an AVC source.

(In other words: CRF 16 with a VC-1 source gives a lower filesize, but lower quality rip than CRF 16 with an AVC source.)

Are VC-1 encoded source files less compressible in x264?

[setarip_old]

Are the VC-1 files interlaced or progressive? (I ask because there are MANY postings at these Doom9 Forums indicating difficulty processing Interlaced VC-1 files)...

[Kurtnoise]

Are you read carefully what the OP said ?

[akupenguin]

Are VC-1 encoded source files less compressible in x264?
There is an effect (http://forum.doom9.org/showthread.php?p=993496#post993496) in that direction. I don't know if this is the main cause of what you're seeing.

[nhakobian]

There is an effect (http://forum.doom9.org/showthread.php?p=993496#post993496) in that direction. I don't know if this is the main cause of what you're seeing.

I had never thought about that in this way, but it makes complete sense! What is even more mind-boggling is that the size of these subspaces changes depending on the encoding settings. What you would need to make "transparent" encodes between two different formats is to enlarge the subspace of the target format to include all of the source's space.

At first I thought it might be as simple as increasing the bitrate, but that doesnt necessarily take into account all of the possible encoding artifacts from the source, which again, I think is related to what you were saying in that post. In fact, it may not be possible to get a 100% transparent encode at ANY bitrate or combination of settings (but it should be possible to get really really close).

[nibus]

VC1 is always tricky to encode, and many times it's impossible to get a transparent encode in a reasonable file size (especially when some of the sources are already so small). One thing that does help sometimes is disabling mbtree altogether, and then raising the CRF to accommodate for the added bitrate that no-mbtree adds.

[buffyangel108]

Thanks for your replies, everyone - very helpful. :-)

[benwaggoner]

Also, beware of VC-1 postprocessing. If you're using the built-in Windows decoder, you should download WMV PowerToy and turn off postprocessing for very clean sources. The more aggressive postprocessing settings can really soften the image, and could thus result in the "lower file size, lower quality at CRF 16" behavior.

And watch out for interlacing, yes. Windows 7 has a pretty good built-in deinterlacer, but it's not as good as a dedicated AVISynth filter.

One good test is to encode at something crazy like CRF 8. If that still looks a lot different than the VC-1 source, you probably have a problem upstream of x264. If you're using AVISynth, open the AVS in VirtualDub and compare that to the straight VC-1 and see if there's a visble degredation there.