How efficient is XviD?

[EnergeticBence]

I was wondering, at what relative bitrate would Xvid acheive the same quality from uncommpressed material as MPEG-2? / When archiving films. /
The only thing I have are some digital home videos to test, but the source isn't good enough to draw conclusion, noise seems to be a problem.

Has anyone done tests in this area?


-EB

[hellgauss]

Xvid is the best mpeg4 codec around

Mpeg2 IS NOT uncompressed. However mpeg4 is 'optimized' for low bitrates, so at 600-1000kbps mpeg4 will be VERY BETTER than mpeg2. At high bitrates (5-6000kbps) mpeg4 is better than mpeg2, but xvid settings should be 'optimized' for high bitrate. Remember also that if you are compressing any kind of video with a lossy codec, the result will loose some quality. So if you are compressing a dvd in xvid, the original will be better than the copy. But this don't means that mpeg2 is better than mpeg4.

[EnergeticBence]

Oh sorry, bad phrasing.
So the Q is: How much smaller is the same quality MPEG-4 compared to MPEG-2; both from the same uncompressed digital source?

[Soulhunter]

Imo XviD can reach the same quality (visually) @ 25% lower bitrates !!!

Note: This comment is only representative for my own eyes... ;)

Bye

[sysKin]

You mean 75% I hope?... ;)

[Soulhunter]

Unfortunately not... :(

Good eyes are the bane of my life !!!

Remember, Im the "1024x576 @ 1x DVD-R" encoder... :D

No matter, most ppl I know say that XviD reaches transparency @ 50% !!!


Bye

[EnergeticBence]

So that would be around 8-10 MB/sec for NTSC material?

( DVD's contain artifacts even if they are around 8.5 MB/sec, good example is Schindler's List )

[Soulhunter]

Originally posted by EnergeticBence

So that would be around 8-10 MB/sec for NTSC material?

It always depends one the content... ;)

The 25% was a "worst-case" limit -> to reach visually lossless quality !!!

And, I was talking about PAL... I dont have much experience with NTSC stuff... :o


Bye

[virus]

Originally posted by Soulhunter
No matter, most ppl I know say that XviD reaches transparency @ 50% !!!
but the original question was not "at what bitrate XviD transparently re-encode MPEG-2 material (like DVDs)"... I think it was "at what bitrate XviD can deliver the same quality as MPEG-2 on uncompressed material".
There's a big difference... ;)
Originally posted by EnergeticBence
I was wondering, at what relative bitrate would Xvid acheive the same quality from uncommpressed material as MPEG-2?

[Soulhunter]

Originally posted by virus

but the original question was not "at what bitrate XviD transparently re-encode MPEG-2 material (like DVDs)"... I think it was "at what bitrate XviD can deliver the same quality as MPEG-2 on uncompressed material".
There's a big difference... ;)

Yeah, I know...

With "transparency" I meant "reproduces 99% (visual) of of the source quality" !!!


Like this...

Uncompressed -> MPEG2-vs-XviD = XviD performs 25% better than MPEG2

MPEG2 source -> MPEG2-vs-XviD = XviD performs 25% better than MPEG2

MPEG4 source -> MPEG2-vs-XviD = XviD performs 25% better than MPEG2


Means...

In all cases XviD could reach transparency @ 25% lower bitrate than MPEG2 would need !!!

This was more about -> How much more efficient is XviD @ high bitrates than MPEG2... ;)


Bye

[chilledoutuk]

i agree with soulhunters 25% for a minimum compression improvement.

Most people forget that they are encoding with a frame size that is normally much lower than that of the anamorphic source video.

People easily forget that MPEG-2 is still a very good video codec and with encoders like cinemacraft can achieve good quality at considerably low bitrates.

When you encode a video whilst retaining the orginal framesize you then see the improvements of MPEG-4's compression over MPEG-2's is a little over rated.

Although it always depends on the source video that you are encoding from as that 25% improvement can go to about 50% if your lucky but dont count on it.

I Have been encoding videos lately with a larger framesize of 1024*576 perfect 16:9 and i have to say that i was impressed by the performance of nero digitals MPEG-4 ASP codec when encoding hi-res video like this.
I do believe that the option to set the MV range to -1024 to 1023.5 has enabled recode to compress higher resolution video better than xvid with its mv range limited to -128 to 127.5

[Teegedeck]

XviD reaches transparency to an MPEG-2 source at about 40-45%; and I have a discerning eye and a sinfully good 19'' Eizo TFT.

Using SixOfNine at about quant=~3.5 to 4, b-frames (defaults), AQ, GMC, Trellis, QuarterPel, VHQ=4 etc. etc.

For the less discerning eye, SixOfNine-HVS at quant=5 or 6 might be enough to reach transparency, ergo something like 30%.

MPEG-4 ASP's capabilities are by no way overrated; nothing one can do against wrong usage of these capabilities though. Ateme's MPEG AVC codec is very good and reaches the same compressibility as XviD at transparency level without much tuning. (And this does mean Ateme's codec is very good indeed.But it still doesn't reach any considerable bitrate savings at that quality level. Its main area seem to be low bitrates where it can make use of its inbuilt deblocking/loopfiltering.

Originally posted by Soulhunter
Good eyes are the bane of my life !!!

Remember, Im the "1024x576 @ 1x DVD-R" encoder... :D


I take it, you know enough about encoding to be aware that upsizing an anamorphic picture (or any form of resizing for that matter) is NOT allowed if you want to compare compressibility at a level of transparency? Or if you didn't mean that, do you use the 'wrong' matrix? ;)

[Soulhunter]

Teegedeck,

I bet you know that I wrote this comment only to emphasize my quality necessities !?!

Sure, my 25% assumption was based on a simple 1:1 (re)encode scenario... ;)


And I think you have forgotten to add a "imo" somewhere in the 1st sentence of ya last reply !!!

Imo its possible to see a difference between a MPEG2 encode and a 55% smaller XviD one... :p


Btw, could you plz have a look @ this (http://forum.doom9.org/showthread.php?s=&postid=561301) ???


Tia n' Bye

[Teegedeck]

Yes, the XviD encode looks sharper! :D
...if you use QP.

Well, enough fooling around, of course I should always say 'to my eyes' because my eyes they are. Also, everyone has different preferences and techniques. For example, some people will insist on filtering and sharpening their picture because they cannot live with the original's look and consequently reach a lower compressibility. Some people don't use B-frames. Or adaptive quantization. And so on and so on.

Also, there are bound to be differences because, well, at least I don't have uncompressed source material. Lastly, MPEG-4 has quarterpel, which MPEG-2 hasn't, and for DCT-precision it is the other way round. A different 'look' might be a result. Though I feel quarterpel pretty much reaches something for MPEG-4 that DC-precision reaches for MPEG-2.

But still, if the source is of high quality, I say: 40%
This means a 60% higher efficiency!!

Edit: You weren't by any chance using the aptly named 'EQM V3UHR' matrix for your compressibility/transparency tests, were you?

[Soulhunter]

Originally posted by Teegedeck

Edit: You weren't by any chance using the aptly named 'EQM V3UHR' matrix for your compressibility/transparency tests, were you?

Ehm, you mean coz Ive added this link @ the bottom of my last reply ???


Nope, not really... :D

I want that you have a look @ this "forum database error" stuff !!!

Because you are a mod of the XviD forum and so... ;)


Bye

[Teegedeck]

I thought, 'well, one of the usual browser-cache problems', and I usually don't pay much attention to those. (Yawn...) But OK, I'm gonna have another look at it then. Edit: Done.

[Sagittaire]

MPEG2 vs MPEG4 ASP: make PSNR/SSIM test for bitrate equivalence ... :devil:

For me visualy bitrate/2 with DVD resolution ...

[Soulhunter]

Originally posted by Sagittaire

MPEG2 vs MPEG4 ASP: make PSNR/SSIM test for bitrate equivalence ... :devil:

You mean the percentual difference from a "perfect" PSNR score ???

Nah, just kidding... :D


Bye