Could VBR ever be better than CBR??

[vitualis]

This is not a flame DJRumpy, but those links have been provided before and they also don't contain any information useful in pertaining to the specific point of contention.

Regards.

[DJRumpy]

Not to start this whole thing up again (too late...), but I believe the second link is very pertinent. They touch upon the question regarding quality of an encode, maximum bitrate, motion detection, fade detection, etc.

"For VBR to visually outperform CBR, a mix of "easy" scenes and "difficult" scenes is always required. If all scenes were the same (easy or difficult), the VBR results would be equal to those for CBR."

This would indicate that if one is using VBR for encoding, and the max bitrage was set to 2500, also, using CBR with a bitrate of 2500, and the entire clip required a 2500 Mb/sec or higher bitrate, then the CBR, and VBR encodings would look identical, with no difference in quality, because the VBR would automatically adjust the output to the necessary bitrate, or if the necessary bitrate was not possible due to a defined limit (2500 in this case), then it would produce output at the max bitrate.

Basically both would produce a 2500 Mb/s mpg, regardless of min/average settings.

[vitualis]

This would indicate that if one is using VBR for encoding, and the max bitrage was set to 2500, also, using CBR with a bitrate of 2500, and the entire clip required a 2500 Mb/sec or higher bitrate, then the CBR, and VBR encodings would look identical, with no difference in quality, because the VBR would automatically adjust the output to the necessary bitrate, or if the necessary bitrate was not possible due to a defined limit (2500 in this case), then it would produce output at the max bitrate.

This is not necessarily true and I believe to be not true. As adam has previously highlighted, just because the bitrate used is identical does not imply identical quality. If each subsequent encoding pass in VBR encoding improves the motion search precision (or vector estimation, etc.) -- and I believe it does -- then multipass encoding could yield better quality.

In a way, it would be similar to changing the "motion search precision" from "very fast" to say "very slow" in TMPGEnc. Same bitrate, but you get better quality because the encoder spent longer trying to find the optimum results.

Furthermore, it seems to me that the second link (which has some fantastic info BTW) is talking about the more usual type of comparison between CBR and VBR; that is, where the average bitrates are the same. The bit about the "fade out" is important as you've highlighted. On the second/subsequent passes, the article demonstrated how the mpeg encoder can "do better" for those parts of video when compared to a single pass/CBR encoding. Again, this implies that for the same bitrate (and we are talking about even for the same maximum bitrate), VBR encoder could hypothetically look better than CBR.

I emphasise again, however, that there will be very little difference. If you really are tossing up between a CBR at 2500 kbit/s vs. multi-pass VBR where the average can be the same, you should definitely go CBR all the way.

Regards.

[DJRumpy]

I guess I didn't really describe my position on this very well (understandable given the size of this post... 8) . I agree with you. The second pass would only improve the encoding. The document states that the second pass does not limit itself to bitrate allocation (this was a key question for this post), it also takes advantage of the second pass, evaluating motion, fade, quantization, pixel luminance, etc.

The only reason I choose VBR as a superior encoding method, is because it can duplicate CBR, bit for bit, and produce a smaller file because it's more efficient. I also agree , that if your mpg is small enough to fit on an svcd, than the extra time isn't worth the bother, making CBR the obvious choice.

The confusion over quality seems to come in over the AVERAGE and MINIMUM setting in 2-pass vbr. The setting itself doesn't limit the output in any way. If 2524 Mb/sec is what is called for, than that's what will be supplied to the stream. ( could we possibly beat this dead horse any longer? )

[adam]

The confusion over quality seems to come in over the AVERAGE and MINIMUM setting in 2-pass vbr. The setting itself doesn't limit the output in any way. If 2524 Mb/sec is what is called for, than that's what will be supplied to the stream. ( could we possibly beat this dead horse any longer? )

No not at all. Of course the avg bitrate can limit the output, by definition it dictates the output. If the source "requires" 2524 throughout the entire movie and you set your avg bitrate to 2000kbits then tough, you will not get 2524 in every scene. The bitrate the encoder allocates to each second is completely relative to your set average. The encoder will basically attempt to find the least bitrate intensive scenes and lower the bitrate there for the sake of other more bitrate intensive scenes, even if quality degragation occurs in these particular areas as a result. It has to lower the bitrate somewhere to achieve the average set, so think of it as the lesser of 10 million evils. It basically tries to get the most even level of quality overall. VBR attempts to do the best with what you give it but it won't override your settings because the source is more difficult than you predicted. If you set your average to 2000kbits than the total amount of bits divided by your total amount of seconds MUST equal 2000kbits. If it doesnt then you need to get yourself a new encoder.

As for the min setting, yes this does limit your output as well. For instance use a min setting of 1000, an average of 1000, and a max of 9000 and see what you get. You will get what is essentially a CBR encode at 1000kbits. If the bitrate can never drop below 1000 than it can never rise above it either otherwise it can never average out to 1000, just do the math. The min and the max setting create the interval for which bitrate is selected, but all bitrate allocated must average out to your avg bitrate setting.

In asking whether multipass vbr can exceed CBR in quality, where the max bitrate is the same for both, I think you just need to ask yourself a few simple questions.

1) Do extra passes aid in bitrate allocation, at least up to a point?

I think anyone will agree the answer is yes. This is a documented fact.

2) Is there any point in an encode where the VBR and the CBR clip will be limited to the same amount of bitrate?

Yes anytime the VBR clip allocates the maximum amount allowed, theoretically this will occur anytime there is lots of movement, assuming there are points where bitrate can be spared which any movie should have.

So as far as I can tell it all comes down to one question...

Does the encoder use the information gained in each pass to distribute the bits only to each second of video, or does it use this information to also distribute the bits to each frame within each second?

Obviously you can tell that I've already made up my mind, and logic would certainly dictate the latter but just consider what multipass encoding does. If the encoder only analyzed each second of film than I could see how it would only use this information to distribute the bits to each second, and just allow each frame to fend for itself so to speak. But during each pass the encoder gathers information regarding each FRAME. Why should'nt it use this information to distribute bits among those frames, regardless of how many bits were allocated to that second's worth?

So during high motion scenes multipass VBR and CBR will both be limited to, say 2500kbits for that single second. The VBR encode will be able to distribute those bits among the ~30 frames (ntsc) in a more intelligent and efficient manner. If you agree that more intelligent bitrate allocation translates into better quality than you must therefore conclude that multipass VBR can potentially achieve better quality than CBR, even when both are limited to the same max bitrate.

In this light, if multipass VBR can achieve equal quality as CBR in a smaller filesize, than it theoretically can achieve higher quality in an equal filesize, because even though they both have the same amount of bits per second they do not have the same distribution of bits per frame.

In the end the best way to prove this to yourself is to make a test encode. Use a low CBR setting like 1000 and see if the multipass VBR clip with max 1000 looks better. I know I can see a difference.

One thing we can all agree on is that in a situation like this you should still just use CBR.

[12Boiler34]

I read something interesting in this thread, so I would like to join the discussion (actually I have some questions)

Someone mentioned, that it would be possible to increase the max Bitrate in a VBR project up to 3000 kbits. Would such files really produce better quality and would these files be playable on a DVD Player that supports (normal) SVCD?

I also became aware, that the min Bitrate should not be higher than 500 kbits. In a Newbie Guide somewhere else (don't remember where) I read, that the min Bitrate is found out this way:

1150 kbits minus the Audio Bitrate (example 160 kbits). Here, the min Bitrate would be 990 kbits.

Can anybody tell me what the correct way is (with the intention to get the best possible quality)

Thanks a lot

[DJRumpy]

The maximum bitrate for a standard SVCD is 2524Mb/second. If you exceed that, then your taking your chances. What you would end up with would be better quality, but it would also violate the SVCD standard (giving you an xSVCD). Some players support the X standards (or lack thereof), but the majority do not. Make sure you do a small test with a CD-RW first. If your DVD standalone doesn't support CD-RW, then you'll just have to chance a coaster.

[adam]

12Boiler34 I don't know what that guide was referring to but it was not the min bitrate to use in VBR encoding. Your min and max setting create an interval from which the encoder selects bitrate to allocate to any given scene, and of course all bitrate allocated must average out to the amount you specify. The bigger this interval the better the quality, so this means the lower the min the better and the higher the max the better.

Of course there are other things you have to take into consideration. Though a min of 0 is theoretically the best value to use, no encoder is perfect and some may allocate too little bitrate during low motion scenes, though I never see this happen. Also some dvd players may stutter when the bitrate drops too low, so its best to use a min value of between 300-500, or higher depending on the limitations of your dvd player. Using a higher min bitrate will actually decrease overall quality.

As for the max bitrate settting, the max video bitrate allowed for SVCD is actually 2600kbits, not 2524kbits. (definitely not 2524Mbs ) However the total bitrate (audio + video + any overhead) is around 2778kbits.

So if you factor in an audio bitrate of 224kbits you get a reasonable max of around 2554 (2778-224=2554.) Factoring overhead into the equation, a reasonable max for SVCD is around 2500kbits.

As DJRumpy said, a higher max bitrate will increase quality up to a point but if you go outside of the standard then playback will not be guaranteed, though many players will not have any problems.

[12Boiler34]

I see, thanks for the info, Adam

[boathouse]

Thanks everyone, that all makes my original post much clearer

Hi Boathouse

I quess your thread was really messed up

Since this rather heated & and alot of the time senseless debate I am very very sure your conclusion at the start was corect. I did a lot of reading.......

If my VBR encoding has a maximum bitrate setting of 2520, and my CBR setting is also 2520, I can't see how VBR can be better, since with CBR you're giving it the maximum bitrate all the time.

You were 100% corect VBR can not beat CBR at these settings. Never Ever. The strange ramblings about motion search precesion geting better with each pass making any significant impact on the encode were really not true.(at least for Tmpgenc & most other 2-pass encoders)


For Example from another page...........

OK, here we go: The second pass is for optimisation of the Bitrate for each GOP and each frame. The motion search will be done in the first pass and in the second pass, and there is no chance to change it or optimize it in the second pass. All Quantisation Levels, calculated in the First Pass, are the result of the Pictures and the Motion. Like i said before: if you change the motion search precission, you change every thing.
But, if the encoder is able to use a higher Bitrate for a GOP, he isn't forced to use a high Quantisation Level for motion compensated Macroblocks. This has a very familiar effekt compared to optimized Motion Search.

If you do 2pass with equal AVG and Max, you are doing the same like 1pass, but in the double time. You need a clear difference between AVG, Min and Max do get good results with 2pass.

And Again: The Bitrate control mode has nothing to do with die Motion Serach accuracy.

Cheers
Matt

[BJ_M]

......................
Of course there are other things you have to take into consideration. Though a min of 0 is theoretically the best value to use, no encoder is perfect and some may allocate too little bitrate during low motion scenes, though I never see this happen. Also some dvd players may stutter when the bitrate drops too low, so its best to use a min value of between 300-500, or higher depending on the limitations of your dvd player. Using a higher min bitrate will actually decrease overall quality.............

the value of 0 has both changed over time for different versions of TMPGEnc and also depending on what encoder you are using ..

on any version of TMPGEnc after 2.0 a value of 0 as min in VBR encoding causes problems with (near or absolute) blacks in dark scenes - therefore i have suggsted to set this at apx .. 300 for svcd and 2000 for dvd ..
with padding ...

of course each should determine on thier own thier best settings for that film.

CCE and MC seem to handle a lower min better (as did tmpgenc 1.2j) - though in the long run you can produce very close results between most of them (re: black background)..


interisting discussion on CBR vs. VBR and i dont really know who is right (nothing like stradlng the fence) - all i can add is my own experiance ..

low bit rate material (under ~6 meg/s) we think VBR 2 pass looks better (or CQ acually) , under 2meg/s we use 3 pass to better dist. the few bits that are avaiable
very high bit rate material 19meg/s and up is always encoded CBR

a LOT of dvd material is produced at 7-8.8 meg/s and is almost done at CBR ... mainly for time and it looks good.

hardware encoders (the better ones) can do both VBR and CBR ..... but depending on the encoder on what looks better
..

[vitualis]

You were 100% corect VBR can not beat CBR at these settings. Never Ever. The strange ramblings about motion search precesion geting better with each pass making any significant impact on the encode were really not true.

As I stated before, the likely impact is small (indeed, probably completely unnoticable), but "NEVER EVER" as quoted by you is definitely incorrect.

The motion search precision does improve with each subsequent pass and the second link by DJRumpy http://researchweb.watson.ibm.com/journal/rd/434/westerink.html has an example that highlights this clearly.

The second and subsequent passes does MORE than just bitrate allocation. Otherwise, each pass wouldn't take as long as the first.

This is the take home message.

The second and subsequent passes does MORE than just bitrate allocation.

Regards.

[BJ_M]

so what does it do ?

Hi Vitualis

I also read the link you gave +many others, and I still stand by what I say. I guess we just have to agree to dissagree

also lets just talk about Tmpgenc just for an example, try and simplify the discussion........I would be interested in your feedback on this.....


1) The Motion Search Settings are used for both passes.

2) In the First Pass, TMPGEnc looks what Bitrate is necessary for a given Q-Factor. (motion search is a factor in the q-factor)

3) In the second pass it changes the Q-Factor dynamicly to get the given
AVG-Bitrate.


4)Motion Search is the process for recognize Motion in the Movie. This describes, how to search motion. The Ammount of found Motion is important for things like Quantisation, so you get other Quantisation-Levels if you use an other Search Level. Theoretical it is posible, to look for Motion Optimisations, but NOT in only 2passes.
Maybe a Encoder can look for Motion in the First pass, then try to verify this in a second pass. A third pass will look for the Quantisation-Levels and a fourth one will do the Encoding-Job.???


you wrote

The second and subsequent passes does MORE than just bitrate allocation. Otherwise, each pass wouldn't take as long as the first.

Well actually the second pass in Tmpgenc 2.58 seems much faster on my PC, I´ll time it on the next encode I make to be 100% sure.

Cheers
Matt

[mats.hogberg]

In every discussion on quality, we must take file size into account - if we don't then there's no point!
The question should be - What gives the best result on a given file size?
As far as I can see, encoding VBR at average bitrate Y will always give equal or better results than encoding CBR at bitrate Y.
Take the extreme - a movie consisting of a still. Using VBR, only the first frame would be needed, as all the rest are the same.
Thus, using VBR, the every bit in the resulting file could be devoted to this first frame (as all the rest of the frames are the same as the first).
Using CBR, each frame would get it's share if bits, to no use.
This is of course theoretical, but I hope you see my point.
The other extreme is where each frame is completely different from the previous. In this case, VBR and CBR will give you the same output (even if CBR will get you there quicker, but that's beside the point).

/Mats

[adam]

The encode mode, VBR or CBR, just controls how much bitrate is allocated every second. So at the same settings VBR and CBR would be equal, but D_Head you said it yourself, we do not encode in all I-Frames
Different frames within each second will require different amounts of bitrate. How this bitrate is distributed is determined by the motion search algorithm of the encoder. The higher the level of motion searching the more exaustive an effort the encoder uses in searching motion, and thus the better allocated the bits are per frame....even though the amount of bits per SECOND is the same. So it should be very easy to see how two encodes can have the exact same min, avg, and max bitrate and yet one can still look better than the other. If you disagree do what has been suggested numerous times already, encode two clips in CBR at the same settings, one at highest motion search and one at lowest. The total bitrate used every second is identical, yet the allocation each second is better in one than the other, hence the higher quality.

The only objection I can see to suggest that x-pass vbr wouldnt produce theoretically higher quality than CBR at same avg and max, would be if you assumed that each subsequent pass didn't improve motion searching. But considering that motion searching is literally the process of gathering information about your source, how could a pre-encoding scan of the source NOT aid in gathering motion search information? The encoder uses this information to better allocate bits per second, this is indisputable and is well documented. Why wouldn't the encoder also use that information to further distribute the bits per frame as well? It would make absolutely no sense not to.

Look D_Head you obviously just don't understand and nothing that I nor Vitualis can say will make you understand. So instead of trying to wrap your head around this concept why don't you just run a test and see for yorself. In TMPGenc make a decent length sample of an encode in CBR at 500kbits with lowest motion search precision. Now do the exact same encode but in 2-pass vbr at min, avg, max 500kbits. The 2-pass VBR should look better.

This discussion is moot. There's no point arguing whether x-pass VBR is better than CBR at same settings because there is no practical reason to ever do this.

Hi Adam

The same questions/concepts I put forward to Vitualis I would also like to discuss with you eg 1-4 in my previous post. And for now lets keep it related tp Tmpgenc 2-pass or any 2-pass VBR encoder for that matter. Your coments/feedback arguments would be welcomed. Who knows maybe we both learn something new?

Cheers
Matt

One last comment to anyone who enters this discussion, please no more, I`m right and youre wrong, you dont understand, etc etc
explain your theories give examples, discuss, debate discuss, discuss, debate & give your point of view,

cheers

[adam]

I'll try to respond to your 4 statements but I don't really see how they are at all relevant to this thread...

1) The Motion Search Settings are used for both passes.

Ok but that doesnt mean that the results of the motion searching can't be improved upon in each pass. This is just how multipass encoding works, it gathers more and more info about the source each time it scans up until it reaches the point where no more information is left to be gained. So if you want to place a purely arbitrary hypothetical comparison between motion searching and extra passes, then say, 1 pass at high motion search precision is equal to 2 passes at low quality motion search precision. In the first case the encoder searches much more diligently to find motion differences but it only has one shot to do it in. In the second case the encoder is not searching as much but it has two tries to improve upon itself. So at the same motion search setting the 2-pass vbr encode should have better motion compensation.

2) In the First Pass, TMPGEnc looks what Bitrate is necessary for a given Q-Factor. (motion search is a factor in the q-factor)

No not at all. In multipass encoding YOU specify how much bitrate to use, this is completey separate from any processes the encoder might perform. The encoder then just tries to use that set bitrate amount as efficiently as possible. The quantization level is the amount of data thrown away and obviously any encoder is going to try to preserve as much of the original source's data as possible, but don't put quantization into the equation, if anything its more of a product rather than a factor. During the first pass, or any subsequent pass, the encoder is gathering as much information as it can about the source so that it can best distribute the bitrate that YOU specify. Better motion search just means that with each pass the encoder is able to gain that much more information about the source and remember motion search precision is an option in an encoder, not a requirement. Other encoders simply use what they determine to be the best middle ground between motion searching and encoding speed. TMPGEnc allows you to customize its motion search algorithm, but its not like its a separate thing from the encoding process. Just like any other aspect of encoding, the more information the encoder has the better job it can do at the given task. Regardless of what your motion search precision is set to in TMPGenc, either way it is going to try to achieve the highest quality, and the lowest Quantization level with whatever information it has.

3) In the second pass it changes the Q-Factor dynamicly to get the given AVG-Bitrate.

Again the Q factor is more of the product of encoding. The encoder adjusts bitrate to achieve the lowest overall Quantization level, and as a result it allows the quantization level to rise or lower from one scene to the next.

4) The Ammount of found Motion is important for things like Quantisation, so you get other Quantisation-Levels if you use an other Search Level. Theoretical it is posible, to look for Motion Optimisations, but NOT in only 2passes. Maybe a Encoder can look for Motion in the First pass, then try to verify this in a second pass. A third pass will look for the Quantisation-Levels and a fourth one will do the Encoding-Job.???

It is not theoretically possible to look for "motion optimizations," it is a fact that occurs everytime you encode something. The encoder analyzes motion from one frame to the next to determine where the bitrate is needed most. This determines how much bitrate gets allocated per second and where those bits get allocated per frame. Yes this happens in both 1 pass and multiple pass encoding but the point is that the more times the encoder scans the source, the more information it has to encode with and that even when given the exact same number of total bits to work with, a more informed encoding mode can produce higher quality results.

Did you know that when encoding in CBR the quality gets better the further into the movie you get? This is because there are more P frames already encoded for the encoder to go back and compare the current P frames to. This is what motion searching is! So if the entire movie has already been encoded once, twice, or three times before doesnt that give the encoder more predicted frames to look at? Can't the encoder therefore allocate bitrate better? The answer is yes.

This is all really very simple. In 1 pass encoding, CBR or VBR, the encoder is gathering information on the fly and applying it to its encoding process. In multipass encoding it gathers information first, and uses this as a guidline when encoding the second time, in which it again gathers even more information about the source.

We all know that at the same avg and max VBR and CBR have the same amount of bits per second. But if the encoder has more information about the source than it can better distribute those bits per frame.

[vitualis]

I don't really have much more to add than what adam has already stated.

As for "motion search precision", I use this term liberally and I don't really mean the specific option as listed in TMPGEnc -- rather, what it does.

adam has already explained this above, but consider we set the "motion search precision" option in TMPGEnc to very fast versus very slow. Why does one look better than the other even though the bitrate settings can be identical?

It is because the encoder spent more time working out the "optimal" (but only optimal as so found by the specific algorithm) encoding settings in the "very slow" option.

When you encode in multipass encoding, this "motion search precision" (the concept, not the TMPGEnc specific option) improves on subsequent passes. If you read DJRumpy's second link, there is a specific example in that article that describes how in a particular circumstance the second pass improves...

Fades
Quite often, the transition between two different scenes is realized not with an abrupt scene cut, but with a fade. This may be done directly between two scenes, or the old scene may first be faded out toward a blank scene, followed by a period of blank pictures, and completed by a fade-in from the blank to the new scene. During fades, MPEG-2 encoders may have problems maintaining coding efficiency, especially if the fade is not very gradual and the pictures in the fade differ too much for efficient prediction. Practical motion estimators are generally not designed to match blocks with different luminances, though features may be similar. An estimator therefore has difficulty identifying the correct motion vectors during fades, resulting in large prediction errors and random motion fields. These random motion fields consume large quantities of bits. Especially for B-pictures, this may be disproportionally large compared to the total number of bits for the picture. Even if the correct motion vector is found, the prediction errors may be larger than normal, resulting in a need for more bits to encode at a certain quality.

It is possible to improve coding efficiency during fades in a second pass. One measure is to force zero motion vectors everywhere. We can thus free bits from coding the motion vectors, and use them for coding the prediction errors more accurately. Note that if there is very little motion, which is often the case at the beginning and end of a scene, we actually are using the "correct" motion vectors, which may help to reduce the visual artifacts. Setting motion vectors to zero is quite a drastic measure, so we wish to be sure that we apply it only to cases of clearly problematic fades.

BTW, I can accept that not all MPEG encoders do multipass encoding in the same way (e.g., TMPGEnc vs. CCE).

Regards.

[vitualis]

In every discussion on quality, we must take file size into account - if we don't then there's no point!
The question should be - What gives the best result on a given file size?
As far as I can see, encoding VBR at average bitrate Y will always give equal or better results than encoding CBR at bitrate Y.

This wasn't necessarily the discussion, and in the discussion at hand, we are assuming that the average bitrate is the same.

BTW, VBR need not necessarily give better results than CBR. I could easier write a VBR encoder that inappropriate gives lots of bitrate to easy to encode scenes and little bitrate to hard to encode scenes. In such a case, the CBR clip would look better no?

What I am trying to highlight here (and doesn't actually have much to do with the thread BTW ) is that what an algorithm "sees" as better quality and what our eyes/brain "sees" as better quality are not always the same.

Furthermore, identical bitrate settings does not equal identical quality. Although the min, max and average bitrates can be the same for two video clips, it is (1) how that bitrate is spread WITHIN the file that is important (e.g., CBR vs. same average bitrate VBR) and (2) and how accurate the "motion search precision" is (e.g., same CBR, difference between "very fast" and "very slow" with TMPGEnc).

Regards.

Vitualis

but consider we set the "motion search precision" option in TMPGEnc to very fast versus very slow. Why does one look better than the other even though the bitrate settings can be identical?

This is one point I have never disagreed with, but what I do disagree with is the percieved effective picture quality gains in the second pass(if infact the encoder actually makes motion optimisations in the second pass and not just bitrate opinisations for the second pass).
Especially if the max & AV bitrates are the same as in this example.

Plus also considering there is really no such thing as a CBR bitstream in reality, and that in reality CBR is really a form of VBR, backs up my point that there is no point encoding 2-pass vbr with the same max & av values.

It is because the encoder spent more time working out the "optimal" (but only optimal as so found by the specific algorithm) encoding settings in the "very slow" option.

Also agree 100% with this -never disagreed with this point

When you encode in multipass encoding, this "motion search precision" (the concept, not the TMPGEnc specific option) improves on subsequent passes. If you read DJRumpy's second link, there is a specific example in that article that describes how in a particular circumstance the second pass improves...

I think I have read this link before, was that the IBM link, I think this was a theoretical encoder, but I will read this again

so again going back to the original post by boathouse

If my VBR encoding has a maximum bitrate setting of 2520, and my CBR setting is also 2520, I can't see how VBR can be better, since with CBR you're giving it the maximum bitrate all the time.

It wont be

Cheers

[adam]

This is one point I have never disagreed with, but what I do disagree with is the percieved effective picture quality gains in the second pass(if infact the encoder actually makes motion optimisations in the second pass and not just bitrate opinisations for the second pass).
Especially if the max & AV bitrates are the same as in this example.

Motion "optimizations" and bitrate optimizations are the SAME thing! The motion of the picture looks better because the encoder allocates bitrate per frame more intelligently. So you are agreeing that the 2nd pass increases bitrate optimization, ok then you are admitting that x-pass vbr "can" produce higher quality than CBR at the same settings, good enough for me.

Plus also considering there is really no such thing as a CBR bitstream in reality, and that in reality CBR is really a form of VBR, backs up my point that there is no point encoding 2-pass vbr with the same max & av values.

No this flys directly in the face of your stance. You are contradicting everything you have said up to this point. If CBR really isnt truly constant then why not? If you look at its bitrate allocation in a bitrate viewer the line is still reletively straight is it not? So how is it not constant? Because it fluctuates per frame. So if both encoders have the same amount of bitrate per second, they still won't have the same amount of bitrate per frame. You have already stated numerous times that multiple passes will increase in bitrate allocation, so if both encodes are essentially done in VBR mode, then the multipass version will allocate bitrate better per frame. Exactly who's side are you on here?

Adam Wrote

Motion "optimizations" and bitrate optimizations are the SAME thing!

This is 100% not true ADAM -Motion optimisation(vectors) and bitrate optimisations are 2 very very different things!!

The motion of the picture looks better because the encoder allocates bitrate per frame more intelligently. So you are agreeing that the 2nd pass increases bitrate optimization, ok then you are admitting that x-pass vbr "can" produce higher quality than CBR at the same settings, good enough for me.

Up to your old tricks Adam twisting the facts

No this flys directly in the face of your stance. You are contradicting everything you have said up to this point

No Adam not true!! read the thread again...


.

If CBR really isnt truly constant then why not?

If you had read just once anything in depth about Mpeg encoding you wouldnt ask me this 8)

If you look at its bitrate allocation in a bitrate viewer the line is still reletively straight is it not?

Wrong again ADAM look in bitrate viwer again

So how is it not constant? Because it fluctuates per frame. So if both encoders have the same amount of bitrate per second, they still won't have the same amount of bitrate per frame. You have already stated numerous times that multiple passes will increase in bitrate allocation, so if both encodes are essentially done in VBR mode, then the multipass version will allocate bitrate better per frame. Exactly who's side are you on here?

Its not about sides ADAM or whos right and wrong, its about discussion!

[adam]

I can't make it any simpler and if you still don't understand than you never will. This discussion is pointless because you are the only one still posting in this thread and you refuse to listen, so lets do what you suggested and agree to disagree.

I am all for forgoing arguments but this is not possible when people like you jump into a civil discussion and turn it into a debate by making inflamatory blanket statements that could never be proven even if you had evidence, which you don't.

I have two words of advice for you. Avoid using phrases like "Never Ever," and do your research before posting inflamatory remarks. Is it not relevant that only about a week ago you posted, "What is the difference between VBR and CBR and which one is better?" and now you are drawing definite conclusions based on your "research"? This thread is just plain silly.

[adam]

If CBR really isnt truly constant then why not?

If you had read just once anything in depth about Mpeg encoding you wouldnt ask me this 8)

Hello? McFly??? I wasn't asking you that question, it was obviously rhetorical because I answered it myself in the very next sentence!

[vitualis]

If my VBR encoding has a maximum bitrate setting of 2520, and my CBR setting is also 2520, I can't see how VBR can be better, since with CBR you're giving it the maximum bitrate all the time.

It wont be

Okay, you've just contradicted the logical conclusion of everything you've agreed with in regards to my last post!

I'll make the argument clearer...

It is possible to improve coding efficiency during fades in a second pass. One measure is to force zero motion vectors everywhere. (from the IBM article)

In this particular example (which is separate from simply bitrate allocation), motion vectors have been more accurate determined but which requires the presence of a previous pass (i.e., 2-pass or more).

That is, even for the same bitrate, the second pass is more accurate / presumptive higher quality.

That is, in such a situation (the article specifically described fades here), a multipass VBR (with max. and average the same as the CBR) encode will have better quality as compared to the CBR encode.

This is what I was trying to get at. Multipass encoding will increase the "motion search precision" with subsequent passes.

As I've stated every time though, the difference with be minor so if you are ever in this situation, choose CBR!

Regards.

[BJ_M]

several encoders are coming out with 2 pass CBR encoding as an upgrade option which bears somewhat the fact that more than one pass must be good for something ...
ive got one coming next week for a demo in fact.

I think the only thing where I have contrdicted myself is when I said I will not post again.!

back again to the original ?

If my VBR encoding has a maximum bitrate setting of 2520, and my CBR setting is also 2520, I can't see how VBR can be better, since with CBR you're giving it the maximum bitrate all the time.

Vitualis- I really do not think I have contrdicted my self here, I still think with the above conditions the VBR will not be better than the CBR encode.

Motion search is an option in CBR & VBR.- I dont why you keep stating why there is a difference between very fast & very slow.-this is obvious!

Now I´ll try and make myself a little clearer..........
I doubt very much any encoder can optimse the motion vectors in 2-passes you would need a 3rd or a 4th pass to re-quantise and re-alocate the bits-does anyone agree with me hear?

and then this will not work properely for the conditions in this thread. eg same max and AV.

BJM-A multipass CBR encoder will truly be interesting, and especially the method they use .

Adam??????????????