Why is 2 pass VBR better?

[thebach]

I make SVCD's and use single pass VBR when copying DVD's, and the picture I get is almost identical to the original DVD. So why should I double the encoding time and use 2 pass VBR?

[PimpInDaHouse]

I think 2 pass VBR sucks when I used it the movie got blocks all over when there was alot of movement. So I use CBR because I think it looks better.

[nsdn]

Depends if u want to squeeze more video on a disk or not.....I have managed to get 70 minutes of SVCD on a 90 min CDR using 4 pass VBR using CCE....although admitedly the film was in 2.35:1 format (very wide screen) it had no artifacts at all......the more passes u have the better the encoder is at allocating bit rate to scenes that require it. Multi pass VBR looks at the WHOLE movie ...then allocates bit rate....so your average bit rate can be much lower ...average bit rate of course is what determins how much movie u can get on the disk.

multi pass VBR does not make for a "better" encode....it just uses space more efficiently.

[thebach]

So 2 pass doesn't help picture quality just space?

[nsdn]

hmmmm i should be more care in making statements like .....multi pass does not make for a better encode.....in fact it does.....what i was trying to say....is for a lower average bit rate u can get the SAME quality...hence you CAN squish it down.....of course if u more accurately allocate where the bitrate is needed then....the resulting quality has to be higher....but if u set your single pass bit rate high enough then....there will not be any visible differance between the two........

SOOO say u have a video that just fits on a disk using single pass VBR....multipass VBR with the same average bitrate as the single pass VBR produces the SAME file size right?...the multipass VBR will have much better quality

NB if your source is crap....VBR don't work real well because it ends up allocating the average bit rate everywhere cuz it thinks the crap is info....

[iparout]

Hi there,
well, just finished reading a Video formats' comparison in a local magazine. I'll make this very simple :

-The number determines picture quality
-The movie was ripped off a DVD, so naturally the DVD picture quality is the point of reference, hence scoring a "10"
-All other scores are in comparison with the DVD quality
-The several formats were playbacked on a standalone DVD player.


DVD ------------------ 10
SVCD 2500 kbps(2 pass)- 9,1
SVCD 1600 kbps (2pass) - 8,8
SVCD (quality 73%)------ 8,3
SVCD 1150 kbps (2 pass) 6,8
VHS ------------------ 8,8
VCD ------------------ 6,5

Numbers speak for themselves ...

Hope I helped.

[vitualis]

The problem with that review is that it doesn't compare multipass-VBR encoding with CBR or single pass VBR encoding.

2500 kbit/s doesn't shouldn't be multipass at all -- this is close to the maximum bitrate for SVCDs so CBR is probably just as good.

SVCD (quality 73%) is meaningless because an average bitrate isn't given (there is no means of comparison).

I find the VHS score of 8.8 very dodgy... It would suggest to me, at least that whoever did the encoding didn't do it properly.

Furthermore, was this a doubled-blinded test? And how many subjects were involved? Unfortunately, quality tests are ultimately subjective and usually small reviews like this can be quite skewed in their result.

Thanks for the post. It was an interesting report.

Regards.

[vlk3]

Maximum SVCD bitrate is 3500 kb/s for 90% of players

Then, made SVCD using minimal bitrate 300, average 1800, maximal 3300

Compare quality with CBR

Do you think, that CBR with 1500 kb/s can compare to high motion scenes with 3300 kb/s ? Whole movie have the same size, of course But quality is MUCH BETTER

--------> CBR SUXX (using VBR wirh 3300 as maximal is enough for Aliens (all bellow 3000 is blocky in rain or flamethrower scenes)). Now, make CBR ! Shit, terrible blocky (same scenes)

[homerpez]

CBR, VBR, blah blah.

I'm on the 2-Pass (or more passes) side. It does, when space counts, make a DRASTICALLY better-looking product, at a given bitrate.

I'm comparing apples to apples. If I need a movie to fit on a CD, I could use (for example) 1000 bitrate CBR, or I could average it out to 1000 using VBR.

With CBR, the picture will never get better, or worse, than 1000 bitrate-quality. We all know it sucks.

With VBR, the picture will rarely actually be 1000 bitrate. On mine, it goes anywhere from 500 to 2300 bitrate! This means, on a scene that badly needs more data allocated (high-action scene), it's available. Where CBR would just generate blocks and artifacts.

VBR is best when it is done 2-Pass (though now I've been trying 3-passes in Cinemacraft encoder) and I don't think CQ_VBR does a good job at properly allocating bits.

It is only by 3-Pass VBR I got a 90-minute movie on ONE 80-minute CD! IN SVCD RESOLUTION! THAT ON A TV, I CAN'T TELL THE DIFFERENCE BETWEEN THAT AND THE DVD

(I'll take it upon myself to stop tooting my own horn).

<font size=-1>[ This Message was edited by: homerpez on 2001-08-30 04:48:01 ]</font>

[vitualis]

<TABLE BORDER=0 ALIGN=CENTER WIDTH=85%><TR><TD><font size=-1>Quote:</font><HR size=1 color=black></TD></TR><TR><TD><FONT SIZE=-1><BLOCKQUOTE>
On 2001-08-30 02:13:33, vlk3 wrote:
Maximum SVCD bitrate is 3500 kb/s for 90% of players</BLOCKQUOTE></FONT></TD></TR><TR><TD><HR size=1 color=black></TD></TR></TABLE>

Completely irrelevant. We are talking about SVCD specifications here.

2500 kb/s is approx. MAX SVCD video bitrate.

VBR with a max of 2500, and average of 2500 = 2500 CBR.

Regards.

[UsualNoise]

On a related note, I too have found TMPGEnc to make blocky mpeg-2's when using 2-pass VBR with a 2150 avg, 2496 max bitrate. (Blocky only in high motion scenes, though.)

When I tried making CBR mpeg-2's with a 2050 avg bitrate there was no blockiness in high motion scenes. Why is that? Has version 12a (which I use) been improved upon with more recent releases in this regard?

The other annoying thing I've found is that after generating my CBR mpeg-2 at 2050 avg (which is the recommended bitrate for 46 minutes to fit on an 80 minute CD), it doesn't actually fit on 1 CD. While using bbmpeg to multiplex I noticed that the 'max bitrate' found is something silly like 2478 ... shouldn't the max bitrate be the same as the average, or is TMPGEnc screwing something up here? I'm using the 'highest quality' motion setting and have the 'use floating point DCT' checked, but I don't see how this should mess up the bitrate.

When using 2-pass VBR (2150,2496) I have no problems fitting 46 mins on one CD. But then there's the blockiness ...

I'm hoping someone else has seen this or might know what's going on here.

[PimpInDaHouse]

I use CBR with 5000Kb/s and it looks better than 2 pass VBR with an avg 2250 and max of 5000Kb/s.

[Greg12]

Thing about this.

If you have 1 second to do something, you gotta sacrifice quality to save time.

If you have 2 seconds you can achive a higher quality.

[PimpInDaHouse]

I don't get it.

[UsualNoise]

I still haven't figured out why running TMPGEnc using CBR 2050 mpeg2 results in video with the following settings (according to bbmpeg):

Bit rate : 256250 bytes/sec (2050000 bits/sec)
Computed avg rate : 257500 bytes/sec (2060000 bits/sec)
Computed max rate : 324800 bytes/sec (2598400 bits/sec)

Why is the max rate so high? And if I try to force using the avg rate, I get lots of overflows ... any ideas?

[Peter]

Why is 2-pass better?
First check to understand how mpeg works: http://www.digital-digest.com/nickyguides/keyframes.htm

Simple explained:
Imagine you see a newsreader. The only thing moving in the picture is his/her head. The table, plant, etc. is not moving. So you can encode the frames after the key(full) frame with very little info. You need only to describe the difference, which means a low bitrate.
When you have an action scene, you have a lot 'moving' around and you will need more data to describe all movings. You need a high bitrate.
If you just encode with a fixed bitrate, you use bits in low motion scenes, which are not needed. When you encode multi-pass, the encoder will save these bits and give them to high motion scenes. With the fixed bitrate, the encoder has less bits for the high motion and you'll see blocks.

Hope you understand it better now.

[dysfunctional]

hmm sounds like CBR is better? what about Manual CBR?

i've been downloading some movies and when i watch them on tv, it sucks! really blocky and i hate it... end up just turning my tv off.


i'll try encode disk #1 with CBR bitrate 3000 and i'll try 2pass VBR at 3000 bitrate?

maybe it'll work somekind of miracle, any of you tried this yet? some say CBR is working better though? im trying this on the movie Fast and Furious, lots of high speed scenes...

please advise thanks

[adam]

UsualNoise: That is the nature of mpeg encoders. Even using cbr you will still get some bitrate peaks above your bitrate setting.

dysfunctional: cbr is not better. you can get better quality in a smaller filesize with vbr. I think you meant to say manual vbr? multipass vbr is the best encoding method in terms of quality and size. All other forms of vbr encoding (manual vbr, CQ_VBR, CQ) are less efficient versions of multipass encoding designed to cut down on encoding time.