Compressibility test doubt.

[Irshad]

Hi,

I'm trying to encode a DVD(720x480) of about 123 minutes
to a single CD using GK 0.35. The following are the configurations:

Bitrate Tab:
* Audio A - 128kBit/s
* Interleaving & AVI Overhead - Calculate frame overhead is selected.
Audio 1 is set to vbr mp3

Resolution Tab:
* Output Resolution - 544x304. This gives a bits/pixel of 0.165

The Xvid options are as follows:

* QPel
* VHQ1
* Turbo
* Chroma Motion
* Trellis Quantization

All other Xvid options have default values. With these settings I do a compressibility test.
It gives me 114% compression. What does this mean ?

[krisq]

it means you should use higher resolution.

[Irshad]

When I increase the resolution, the bits/(pixel*frame)
decreases below 0.05. Shouldn't this value be near
0.17 for a 1 CD rip ?

[Axed]

Why use calculate frame overhead? You will get undersizing with that selected for Xvid.

The bits/(pixel*frame) is really just a guide for when you do your compressibility test. After that, rely on your results of the test for changes in the resolution.

With that much leeway of compressibility you could try using the mpeg matrix instead of the h264 since the mpeg matrix is not only more bitrate hungry, but sharper and less prone to blocking.

[Irshad]

Did a search in the forums and found that 'Calculate Frame Overhead'is needed only for DivX and not Xvid. Unchecked this option. Changed to eqm_v3ulr_rev3 quant matrix.

Before doing the compressibility test, I set the resolution to 640x368 which gave a bits/(pixel*frame) of 0.103. The compressibility test gave 122% and a bits/(pixel*frame) of 0.084 (the resolution is still 640x368).

When I change the resolution to 704x416, the compressibility is 98.1% and bits/(pixel*frame) is 0.082. It is given in the Xvid guide at the doom9 site that it's not good to encode above a compressibility if 80%. What should I do now ?

[BigDid]

Originally posted by Irshad
...When I change the resolution to 704x416, the compressibility is 98.1% and bits/(pixel*frame) is 0.082. It is given in the Xvid guide at the doom9 site that it's not good to encode above a compressibility if 80%. What should I do now ?
Hi,

Lower your final avi size? raise the mp3 bitrate? or keep an existing ac3 audio,
mix of the two :)

Did

[Axed]

Well use a high quality matrix such as the 6of9-hvs instead. This will take your compressibility down to a reasonable amount. BigDid already said your other options.

[Irshad]

Thanks Did, I'll check it out

[niamh]

-Calculate frame overhead needs to be unchecked for Xvid, but you still need to fill the audio box to get the right bitrate :) (so you did right in that regard)

-so long as the compressibility is under 100% all is good, what the guide means is it's a waste of bitrate for no noticeable improvement.

-If your comp test is over(and close to) 100% nothing bad will happen, your file will only be undersized, as the codec will have spread as much as it could so to speak.

-The well known 0.17/0.25 values are about as meaningful as saying all humans are 5'7'' tall if they're white and 6' if they're black. So forget about it altogether :) The only value you're concerned with is the one in the box at the left of "compressibility test"(the one under "first pass stats").That's how namy bits/pixels there are in the original and you won't be able to go over that.

-Anyway, you have that rare gem, an extremely compressible movie. Lucky you.
I agree with the posts above, keep the original audio /increase the resolution /use 6of9 or other / Disable B-frames /Don't use denoising (whichever or all of them)

Note: If you're going to encode your movie close to 90, 100% compressibility anyway, another tip: in XviD configuration, advanced options, quantization, replace all the min values (that say 1) to 2, for both passes ; values 1 can cause sometimes Xvid to grossly oversize in a desperate bid not to undersize the file in these cases. Of course, your output file might be undersized then ;) (At No Loss of quality)

[Irshad]

Sorry for the late response, I was busy with my exams. Now I've got a lot of time.

I increased the audio quality and a few tweaks here and there. Everything is alright now. I did the encoding using the eqm_v3ulr_rev3 quant matrix. It gave excellent results in this case.

There is one more issue. I have a dvd of 3hrs 15mins duration. Want to encode it using Xvid into 2 CDs. Which is the best quant matrix suitable for the purpose ?

Thanks in advance for all the help and responses.

[Sharktooth]

EQM V3 ULR or Jawor's 1CD... but for obvious reasons i prefer the first:)
Btw the bpp rule is only valid with an "average" compressibility. Since you're getting high compressibility you can safely ignore the bpp rule.