Compressibility of Cube 2

[RexManning]

Hi!

I am currently making a copy of Cube 2: Hypercube and encountered something weird (well, not really weird but kind of surprising, at least to me). The GNot compressibility test returns a value of 0.843 (at 704x400). At a percentage of 0.80 that results in a filesize of 3.7 GB which is way higher than what I expected:
Star Trek: Nemesis (1'51") -> 2.7 GB @ 100% (I know that's overkill, but hey :-)
Cube 2: Hypercube (1'30") -> 4.2 GB @ 100%
That is the OGM with two AC3s muxed, but the difference should be obvious anyway.

So, why is it that C2:H returns such a high compressibility value? I'm not really into video compression but as far as I can see C2:H has not many fast motion scenes plus the background is pretty much always the same (lots of white). Shouldn't that guarantee an at least average compressibility?

I'm using XviD 0.9.1 (Nic's build) with QPel and GMC.

[mf]

1) This belongs in the GKnot forum. Compressibility tests have nothing to do with xvid.
2) Try using a development build, stable is hopelessly outdated.
3) Using qpel might not always be a good choice, and GMC is b0rked.

[HarryM]

Originally posted by RexManning
Hi!

I am currently making a copy of Cube 2: Hypercube and encountered something weird (well, not really weird but kind of surprising, at least to me). The GNot compressibility test returns a value of 0.843 (at 704x400). At a percentage of 0.80 that results in a filesize of 3.7 GB which is way higher than what I expected:
Star Trek: Nemesis (1'51") -> 2.7 GB @ 100% (I know that's overkill, but hey :-)
Cube 2: Hypercube (1'30") -> 4.2 GB @ 100%
That is the OGM with two AC3s muxed, but the difference should be obvious anyway.

So, why is it that C2:H returns such a high compressibility value? I'm not really into video compression but as far as I can see C2:H has not many fast motion scenes plus the background is pretty much always the same (lots of white). Shouldn't that guarantee an at least average compressibility?

I'm using XviD 0.9.1 (Nic's build) with QPel and GMC.

Cube 2: Hypercube hasn't good compressibility. I make it too. This movie is much bright and contrast.

[RexManning]

Thanks. No way to get around it then.

Btw, do you have the english chapter names? My version of this movie is german but I'd like to have the english names. Couldn't find then with google...

[sysKin]

Originally posted by RexManning
I'm using XviD 0.9.1 (Nic's build) with QPel and GMC. Could yuu please tell me where you have it from?

Such build doesn't exist, and can never exist, so I guess someone just made a mistake in naming files on their webpage.

IF you're using unstable dev-api-3, don't use GMC, it only decreaes quality.
The setting is only OK for dev-api-4, and then helps only with VHQ.

Radek

[RexManning]

Hm. In that case I'm probably wrong. It's definately a build by Nic but I guess I've mixed up the version numbers.

[Soulhunter]

@RexManning

Hi, I did 3x Cube2 encodes the last days for testing...

The thing with the brightnes is right... But,
you should not care so much about it, because as you said, the most of the background is uniform white colored...

I did a 1CD and a 2CD encode with RV9 and a 3CD encode with XviD...
All 3 where nice... Also the 1CD RV9 encode ! (not too much blocks...)

The thing is that this movie has a lot of film grain (+ a sharp image)!!!

This noise is "good" (no joke) because it prevents blocks also in uniform colored arears... (Dont remove it !!!)

But to use a soft resizeing method is a good idear !!!

My result was nice...

Test it for your self and post back...

Maybe this helps... ;)

Bye

[PS: This movie was confusing me "Hypercube" n' "4th Dimension":D]

[RexManning]

Thanks for your post. I'm going for maximum quality because I've a bunch of friends who have literally hundreds of movies in poor (or very poor :-) quality. That's why I prefer a high-quality XviD at DVD resolution and AC3 sound. I don't have a 5.1 surround system but I'm planning on getting one as soon as I win the lottery ;-)

So, a good rip would result in a file which is about 4 gigs large, and in that case I could just make a 1:1 DVD copy and watch it in my standalone DVD player. So I guess that's what I'm gonna do.

[silver_cpu]

note: I have not watched this movie, so what I'm about to say may not work.

What I would suggest is to run it through some AVS filters, in order to sort out all of the bitrate-raising material. For instance, since the movie is so bright, you might want to reduce the luminance some (unless it also contains dark scenes which would suffer). Also, if the movie has a lot of film grain, you could try using FluxSmooth or one of the other noise filters in order to reduce this effect, and save largely on data space (you wouldn't believe how much noise in the signal hurts compressibility).

Also, try encoding at a different resolution. If you're using GKnot, try changing the height/width to a multiple of 32 (you can set this with the H-modul and W-modul drop-down boxes). This greatly increases compressibility with Xvid, which seems to be optimized for resolutions of this multiple, and also because this generally reduces macroblock overhead. I've experienced great success with this method of data readuction (which also preserves quality).

You might also want to try using multi-channel OGG/Vorbis, since AC3 takes up a relatively enormous amount of space.

Hope this helps!

[RexManning]

You might also want to try using multi-channel OGG/Vorbis, since AC3 takes up a relatively enormous amount of space
Nope... 5.1 Ogg Vorbis does not support channel coupling yet, it won't compress good enough.

[silver_cpu]

Did you try the different resolution, though? And, have you considered multi-channel AAC?

[Soulhunter]

@RexManning
So, a good rip would result in a file which is about 4 gigs large...

But did you try a 3CD 1/2 DVD rip with XviD ???

Sometimes the comp. test tells different result than your eyes... ;)

Bye