CBR vs. VBR

[ablatt]

Am I right to assume that if disk space is not an issue, CBR will produce better video than VBR at the same maximum rate?

And if the above is true, why is it that fast action is blockier using CBR? Do you want a slower bit rate with fast action?

[g0dZpr0fi7]

Well with CBR the fast action scenes get the same bitrate as the credits do. With VBR the action scenes get more and the credits less. So unless your talking very high bitrates VBR kills CBR.

[ablatt]

Yes, but if the maximum bit rate is the same for CBR and VBR then CBR should have better quality throughout, the way I figure it.

[Guiboche]

No. VBR will ALWAYS be better in quality than CBR. The reason for this is simply because VBR can allocate bits more intelligently than CBR.

For QuickTime/RealPlayer/WindowsMedia (I'm not sure if this works the same for Divx, and MPEG-2. If someone can confirm it would be great). The bit-rate can jump above and below the set bit-rate if it needs to. For example, let's say you give a movie a bit-rate of 1000kbits/sec. Maybe the beginning only needs 500kbits for the first 10 seconds. What does this mean? This means the bit-rate can jump to 1500kbits for the following 10 seconds there after. The final file will have a bit-rate of 1000kbits/sec and be the same size as as CBR file encoded at 1000kbits/sec, but because VBR has allocated all the bits intelligently, your end result in a better quality file.

This is why you will always achieve better quality when using VBR than CBR.

Thanks for your time.

Guiboche

[holocaust]

Yes, but if the maximum bit rate is the same for CBR and VBR then CBR should have better quality throughout, the way I figure it.

thats how i see it too cbr offers better quality than vbr at the same max bitrate ie:
a 50 minute movie (xvcd)at CBR 1900 is by far much higher quality throughout the movie than the same clip using VBR max1900 min 1150 theres no way at the same bitrates vbr kills cbr but 2 pass VBR kills CBR and VBR CQ

[g0dZpr0fi7]

Well lets say I encode a clip at 2000min, 3500av,7000max which is common for me. The same clip encoded at 7000CBR will look as good but thats 2x the bitrate, 2x thesize, 2x the CD's. So my math tells me it takes twice the bitrate with CBR to equal VBR.

[Kdiddy]

you math is wrong, VBR filesize is greatly affected by scene complextity..
IE, more complex movie have larger filesizes than less complex movies done at same VBR settings.

[bravo]

maybe i'm wrong,but it seems like a lot of faking the correct answer,if the max bitrate is 1500,it wont make a difference in the action vbr or cbr,BECAUSE THE SAME BITRATE IS BEING USED RIGHT.i think u confused the average of vbr and not the max.the max is the max and nothing more.if u set the max at 1500 vbr then the most u will get is 1500 not 1600 not 1501 but 1500.please correct me if i'm wrong but it seems like people will get confused with the wrong info.

[kinneera]

VBR is a pretty clear winner in my opinion. It's not necessarily so simple as the bitrate - if you allow an encoder to run two or more passes, it can apply algorithms more intelligently in various ways that can still result in a significantly better picture even at an equivalent bitrate (try a VBR with mix/max/avg all the same sometime).

It also seems like a sort of silly argument to say that if a VBR max is equal to the CBR rate the CBR is somehow better. If the VBR is determining that, for the exact same video, bits can be dropped while maintaining the same relative quality level, then the CBR is really sort of an intangible waste of bits. If the CBR bitrate is set to match the average bitrate of VBR, then VBR will be the winner hands-down.

As I see it, CBR only really serves two functions - data rate control when there is a potential physical limitation to reading that data (ie. 1x CD/DVD-drives), or faster encoding when space is no issue. It is certainly no accident that VBR was chosen for DVD and SVCD - it probably would have been used for VCD too if the technique had been more fully realized at the conception of the VCD standard (and the aforementioned physical limitation wasn't an issue).

[tigerwolf]

just one quick thought.

VBR decides how it handles the rate correct, while cbr is constant. Thereby with vbr you are relying on the fact that the software will correctly allocate the proper bit rate in the proper area, which thereby allows for error. If I have a fast action scene, yet the vbr stumbles and does not allocate the proper bit rate for that scene then you'll get blocks. In the fact that vbr relies on its software to determine the action scenes and bitrates throughout the movie, then vbr quality would be wholly determined by the quality of the software you use. While cbr is constant and thereby you'll always come out with a quality picture and won't get any problems. Where as its highly likely vbr could stumble at the beginning, end, or even middle of an action scene. It could mistake a low in the action as a much lower rate or not pick the rate up quick enough when the action commences. Also say your source is something that had a constant bit rate, how will vbr be able to tell where the action is?

anyway, just a thought.

[kinneera]

It makes for a reasonable theory, but a VBR encoder can drastically increase the bitrate on very short notice if necessary. The odds of a short low-motion scene in the middle of a high motion scene causing problems are extremely low and functionally non-existent for any encoder even remotely worth it's salt. In actuality, I would suspect the reverse outcome is more probable, in that the intervening low-motion scene would residually receive more bits than are really necessary.

Consider the worst case scenario - the whole movie is nonstop high motion. All that will happen is that a single-pass VBR will functionally be reduced to CBR functionality. A multi-pass encoder with an average bitrate set to an equivalent CBR bitrate will likely exploit any reprieves it can find whatsoever in addition to coding code more intelligent motion vectors, yielding very marginally improved video quality. For anything other than the worst case scenario, the benefit of VBR should be much more tangible.

[DVantrease]

I have always used CQ_VBR (Tmpgenc), and had good results. I do bump the 'quality' setting to 100, although I admit I have no idea what this really effects. I have never tried '2-pass', my question is does it literally make 2 passes throughout the whole movie? In other words, would an otherwise 8-hour conversion then take 16 hours using 2-pass? Is it worth it?

[SatStorm]

@DVantrease: Read that: http://www.vcdhelp.com/phpBB/viewtopic.php?t=79765

With tmpgenc 251 plus, if you set CQ_VBR at highest quality mode, you encode a file the same time as setting 2Pass VBR at Time Estimate Search mode.
2Pass VBR is a winner in time/quality/file size! That is because you don't have to set Higher or highest quality mode anymore! Time Estimate search mode just do an amazing job!

[ablatt]

Let me ask this question very clearly then. I have seen mention on this site that CBR causes 'blocking' during high motion scenes which led me to believe you'd want a lower bitrate during those scenes.

If I set CBR to 1500 and use VBR 2 pass with a max of 1500, will there be any situation where VBR will produce a better quality mpg for high motion scenes, given I don't care about disk space?

[droolian01]

Hi there, no 0ffence but i think you need to learn some maths.

frame 1 2 3 4 5 6 7 8 9 10

cbr 1.5k 1.5k 1.5k 1.5k 1.5k 1.5k 1.5k 1.5k 1.5k 1.5k
(kilobits)

vbr 1.5k 1k 2k 2k 1k 1.5k 2k 2k 1k 1k

motion L L H H L L H H L L
(H=HIGH, L=LOW)

therefore whole ten frames = 1.5kbits/s x10 = 15.000 bits total

using vbr lets you spend more bits where you need them (i.e. in high motion scenes, at the expense of low motion scenes IF CBR (1.5K) = AVERAGE VBR

now in the vbr line we have av. vbr=1.5k (same as cbr), but minimum bit rate set at 1.0k, maximun bit rate set at 2.0k.
If you add each number in the vbr line up you get 15,000 bits in total, the same as cbr, but the computer has decided to spend this 15,000 bits inteligently, more bits in faster scenes, therefore more bits in faster scenes = more detail = better looking rips

[droolian01]

Shit, i had that really nicely lined up!
Probably more confusing now!

[droolian01]

CBR = same all the time

VBR = different
MINIMUM=SMALLEST NUMBER
MAXIMUM = BIGGEST NUMBER
AVERAGE = in between MAX and MIN

So if you set up a 2 pass vbr as i did above, with AVERAGE vbr = cbr, then the end filesizes should be about the same.

But if you set up a 2 pass vbr as you originally suggested, with the MAXIMUM vbr = cbr, then the resulting vbr mpeg would be much SMALLER than a cbr mpeg. So you are right, the cbr clip would look better BUT it is an unfair and silly statement because the filesizes ar different now.
When i burn a cd i want to fill it right up, you know, 699mb on a 700mb disc. And the vbr 699mb disc will always look better than the cbr 699mb disc.
n.b. unless you faq it up!

[KoalaBear]

I have seen mention on this site that CBR causes 'blocking' during high motion scenes which led me to believe you'd want a lower bitrate during those scenes.

That's incorrect. 'Blocking' is caused by a lack of bitrate. The more motion a scene contains, the higher the number of bits required to reproduce that motion accurately.

If I set CBR to 1500 and use VBR 2 pass with a max of 1500, will there be any situation where VBR will produce a better quality mpg for high motion scenes, given I don't care about disk space?

No. 1500 is 1500; if a scene requires 3000 to reproduce, you'll not have enough bits either way.

Am I right to assume that if disk space is not an issue, CBR will produce better video than VBR at the same maximum rate?

At [S]VCD bitrates, the difference is negligible.

And if the above is true, why is it that fast action is blockier using CBR? Do you want a slower bit rate with fast action?

CBR is no blockier than VBR per se. It's all about bitrate. If you don't have enough you're going to get blockiness in proportion to that deficit regardless of the bit allocation strategy.

[droolian01]

KoalaBear, of course everything you say is correct, but i feel, reading between the lines, that ablatt is confused about some basic concepts, such as attempting to suggest that cbr and MAX vbr are in any way equvalent or indeed comparable indices. I think (please forgive me if i am wrong) ablat is having difficulty with the concept of arithmetic mean, and talking about bit rate allocation strategies is bound to confuse further.

[Kdiddy]

So you are right, the cbr clip would look better BUT it is an unfair and silly statement because the filesizes ar different now.

I respectfully disagree, if anything their quality will be the same. A properly design encoder will allocate the bits properly to maintain scene quality. Thus, if the scene requires the max bitrate, then it will be given, so there would be any quality difference. Any lack to maintain the quality is problem/issue of the encoder or settings within the encoder, not that of CBR vs. VBR.

[ablatt]

Actually KoolaBear clarified if perfectly.

Your example of VBR vs CBR showed VBR with a max of 2K which is higher than the CBR 1.5K rate. So of course, the VBR encoding can go higher than 1.5K when it needs it.

My question was, would VBR with a max of 1.5K be any better than CBR at 1.5K and I now have my answer.

[KoalaBear]

i feel, reading between the lines, that ablatt is confused about some basic concepts, such as attempting to suggest that cbr and MAX vbr are in any way equvalent or indeed comparable indices.

In absolute terms I think he's correct, whether he's confused about the subject or not. A file encoded VBR at 500/1000/1500 can deliver no better quality than CBR at 1500, although the VBR may be smaller and therefore allow more time per disc.

[bewley]

ablatt,

How I understand it, the answer to your question is that the two videos will look exactly the same.

Since your CBR setting is already set to the max bit rate, VBR will not give you any better (or any worse) quality. Keep in mind that when in CBR mode, the encoder performs compression and if the bit rate needed is LESS than the set amount, it pads the results with null bits. That's to say, if a sequence of frames requires 2000 kbit and you assigned a CBR encoding of 2300 kbit, the encoder compresses the sequence to 2000 kbit and then pads it with the remaining 300 kbits -- wasting a lot of space. VBR would encode the sequence at 2000 kbit and then store the 300 kbit to apply later when needed.

VBR easily wins out, however, when you are trying to achieve an average bit rate that is lower than what your source material requires and when the CBR bit rate is less than the MAX VBR bit rate. For example, if you have a movie that "requires" up to 3000 kbit for an encode without brick noise, a VBR encode of 1200-2300-2500 would look better than a CBR encode of 2300 as the VBR encode can spike up over the average and apply bits when needed.

Lastly, when encoding VBR, unless your DVD player requires it, don't pad your video... It's a waste of space...

Anyway, just as an example... Last week I was playing with this same concept using TMPEGEnc. I made two XVCD's... The first was CBR 2300 kbit and the second was 2-Pass VBR 1200 - 2300 - 2517. The resulting files looked exactly the same when played back on my TV... The only difference was that the CBR file was around 133 MB and the VBR was only 80 MB. That extra 53 MB was nothing more than padding.

Good Luck!

[zzyzzx]

Isn't my DVD player (standalone) going to like CBR encoded disks better than VBR encoded disks? I would think it would be easier on it.

[kinneera]

Nah...as I recall, CBR is actually harder to work with from the decoder's perspective.

[vitualis]

Am I right to assume that if disk space is not an issue, CBR will produce better video than VBR at the same maximum rate?

Back to the original question...

If the max. bitrate for VBR = CBR, then the VBR will either look the same or worse than the CBR clip.

The reason to use VBR rather than CBR is encoder efficiency. Assuming a relatively comfortable max. bitrate, if you compare a VBR clip with the same average bitrate rate as a CBR clip, the VBR clip will usually look much better.

For example, unless the encoder is broken:
1. VBR: min 500 kbit/s, average 2000 kbit/s, max 2500 kbit/s
2. CBR: 2000 kbit/s

For the following two clips, the resultant filesize will be the same as the average bitrate is the same. However, the VBR clip should/will have better video quality due to more efficient distribution of the bitrate.


For the following example:
1. VBR: min 500 kbit/s, average 2500 kbit/s, max. 2500 kbit/s
2. CBR: 2500 kbit/s

VBR offers no benefit over CBR. The filesize is the same as the average bitrate is the same. Theorectically, the VBR clip will look exactly the same as the CBR one except that it simply took longer to encode. Furthermore, if the encoder is faulty, the quality may even be worse.

Regards.

[Craig Tucker]

For the following example:
1. VBR: min 500 kbit/s, average 2500 kbit/s, max. 2500 kbit/s
2. CBR: 2500 kbit/s

For the average bitrate to be the same as the max bitrate the bitrate must be constant at the max bitrate. So surely this is in effect CBR

Craig

[Craig Tucker]

Sorry forgot to end quote
Craig

[vitualis]

Exactly. VBR with the average bitrate = max. bitrate is equivalent to CBR.

However, it will still take longer to encode as software encoders aren't that smart to realise this.

Regards.

[kinneera]

I think theoretically, allowing the encoder (lets assume a reasonably good one here, such as TMPGEnc or CCE) to run two passes will allow it to apply its motion search algorithms twice, which results in more intelligent allocation of the bits, even if there are the same number of them. The resultant improvement is probably not worth the 2x encoding time in general, but in some cases it may be worth it - I tested Run Lola Run (almost nonstop motion, not to mention the animation interludes) with CBR and VBR, and even though the average for VBR was the same as the CBR, it still looked noticeably better.

But again, this is more the exception than the rule.