PDA

View Full Version : CBR V VBR (Multiple Passes)


iceman
15th July 2002, 01:32
If the size was not matter!, which method is better for quality purposes? CBR or VBR (Using say 4 passes the maximum). I have tried and it appears that VBR is better than CBR becuase of pultiple passes. I have read it somewhere in the forum that CBR should give you better results with bigger file sizes.

I am :confused:

Matthew
15th July 2002, 01:50
CBR @ Max of 2500 looks better than 4 pass VBR with a max bitrate of 2500. The passes are simply there to improve the accuracy with which bitrate is allocated. With high-bitrate CBR every scene is getting the max so the passes are unnecessary.

However CBR with 2300 may well be inferior to 4 pass VBR with 2100 average and 2500 max. This is because the scenes that need it will get 2500 whereas they won't with CBR.

iceman
15th July 2002, 02:53
Thank you matthew. I will your ideas into practice. Suggestions from other people are welcomed.

waldok
15th July 2002, 08:36
I'm currently testing an encoding at CBR 2550, just to compare it to my usual VBR (min 300 min avg 1800 max avg 2400 max 2550). . One thing is encoding is much faster (that's important to me), and quality is at the max of course but it will require 3 Cds for most of the movies (I usually like it on 2 Cds).

I'll let you know about the results

Waldok

Bigbucks1959
15th July 2002, 13:20
If I wanted a bitrate of 2500 on my svcd?

Could you tell me what numbers to fill in from left to right on dvd2svcd on the bitrate tab?

And also which boxes should be checked?

I am trying to make a CBR of excellent quality.

Thx,
Phil K.


Originally posted by Matthew
CBR @ Max of 2500 looks better than 4 pass VBR with a max bitrate of 2500. The passes are simply there to improve the accuracy with which bitrate is allocated. With high-bitrate CBR every scene is getting the max so the passes are unnecessary.

However CBR with 2300 may well be inferior to 4 pass VBR with 2100 average and 2500 max. This is because the scenes that need it will get 2500 whereas they won't with CBR.

Bigbucks1959
15th July 2002, 13:34
Am I seeing things ?? Or is CBR about half the time as 4 pass VBR?

Thx,

Phil K.

Matthew
16th July 2002, 00:37
CBR should actually take about 1/5 of the time of 4 pass VBR. Because 4 pass VBR involves making the VAF before the four passes, so it's like 5 passes. CBR is like 1 pass ;) So it should take 1/5 of the time.

I've always filled in max, min and max average with 2500 and left minimum average unchecked when making 2500 CBR rips ;) [And selected CBR in the encoder tab of course].

Bigbucks1959
16th July 2002, 01:36
Thanks for the info.

DO you also uncheck (make .vaf file) whjen you make a CBR..sounds like a stupid question huh?

Do you find the quality of your CBR at 2500 bitrate exceeds the VBR 4 pass?

Thx again

Phil K.


Originally posted by Matthew
CBR should actually take about 1/5 of the time of 4 pass VBR. Because 4 pass VBR involves making the VAF before the four passes, so it's like 5 passes. CBR is like 1 pass ;) So it should take 1/5 of the time.

I've always filled in max, min and max average with 2500 and left minimum average unchecked when making 2500 CBR rips ;) [And selected CBR in the encoder tab of course].

iceman
16th July 2002, 02:01
I have done various tests using both methods. I have used Madonna's Music DVD (PAL) to test the quality. There is one very demanding scene (where she flies like an angel!)in the SVCD conversion process but it can show you the difference quite clearly between the two methods. That clip is "Rain". I have notices that the VBR with 4 Pass (Telecide Deinterlacing) is much better than CBR at 2500 (No deinterlacing). In other parts you can not tell the difference.

I tested CBR at 3500 (SmarDeinterlacing) and the result was identical to VBR with 4 passes (Much bigger file size but 1/5 of encoding time).

I think there is more to than just bitrate here. I am currently testing CBR at 2500 with SmartDeinterlacing. I will post the results if people are interested.

I am not expert, so I request from experts is whether I am on the right track. Does Deinterlacing have effect on final result ? What method of Deinterlacing should I use (Using CBR).

Are there any other settings I can use to improve picture quality in difficult scenes (Using CBR)?

Thank you.

Matthew
16th July 2002, 02:13
I don't uncheck vaf, it's check by default so I left it that way. My rule is if I don't understand it don't touch it :D

CBR at 2500 is better quality than 4 pass VBR with a max bitrate of 2500, it can't be worse [with respect to iceman, it must be the deinterlacing which made the difference]. At worst you will have used too much bitrate on scenes that don't it (e.g. credits), which leads to a larger size than need be. Whether the quality difference is noticeable depends on a number of factors - the nature of the film, the average bitrate you are using for VBR, the aspect ratio, the quality of your tv, the quality of your eyes, etc.

As an example, I'd say that generally a 3 cd rip of a 110 min film @ CBR 2500 should be noticeably better looking than a 2 cd 4 pass VBR rip of the same film.

Typically I don't want more than 50 min on a (80 min) cd for 1.85:1 or 2.35:1 movies (4:3 movies can be hard to encode nicely at all). More than that I'm into into CBR territory. Disc changes don't really concern me either, so if I had a slow machine I wouldn't mind sticking a 90 min film on 3 cds.

iceman
16th July 2002, 02:37
Matthew, considering you have input in 97 posts, do you suggestions for deinterlacing?

Bigbucks1959
16th July 2002, 03:11
THx.
Doing a test svcd right now at 2500 bit rate..I did uncheck vaf though..

Oh well let me see how she looks when she is done in 2 1/2 hours...

My normal SVCD time is roughly 12 hours on a 2 hour movie..I am doing a CBR at 2500 bit rate in roughly 3 1/2 hours..

WOOT WOOT

Phil K.


Originally posted by Matthew
I don't uncheck vaf, it's check by default so I left it that way. My rule is if I don't understand it don't touch it :D

CBR at 2500 is better quality than 4 pass VBR with a max bitrate of 2500, it can't be worse [with respect to iceman, it must be the deinterlacing which made the difference]. At worst you will have used too much bitrate on scenes that don't it (e.g. credits), which leads to a larger size than need be. Whether the quality difference is noticeable depends on a number of factors - the nature of the film, the average bitrate you are using for VBR, the aspect ratio, the quality of your tv, the quality of your eyes, etc.

As an example, I'd say that generally a 3 cd rip of a 110 min film @ CBR 2500 should be noticeably better looking than a 2 cd 4 pass VBR rip of the same film.

Typically I don't want more than 50 min on a (80 min) cd for 1.85:1 or 2.35:1 movies (4:3 movies can be hard to encode nicely at all). More than that I'm into into CBR territory. Disc changes don't really concern me either, so if I had a slow machine I wouldn't mind sticking a 90 min film on 3 cds.

iceman
17th July 2002, 01:02
I have tested CBR at 2500 with SmartDeinterlacing. The results are better than without deinterlacing. Just about the same quality as VBR (4 passes).

I am wondering whether I should use temporalsmoother and anti noise filter (2). Any commnents on this would be appreciated.

pepsimaxx
17th July 2002, 22:56
The usage of temporal smoother and anti-noise filter can sometimes be warranted but I guess it depends on the source.

Also beauty lies in the eye of the beholder. Do a few test encodes with various setting (just a chapter) and then see what u like.

iceman
18th July 2002, 01:58
You are right pepsimaxx, I have done all sort of tests. Here are my conclusions Based on PAL DVD converson):

1. CBR can't beat VBR (4 Passes).

2. Temporalsmoother and anti noise filter (2)don't make any difference.

3. Deinterlacing makes a big difference and you can get close to VBR (4 Passes).

I would to hear comments from other users.

adam
18th July 2002, 07:07
As far as bitrate allocation is concerned, if the CBR level is equal to your max bitrate in x-pass vbr than CBR can never be worse than x-pass vbr. In any given scene the VBR encoded file "may" have as much bitrate allocated as the CBR clip, but it will never have more. Of course this is assuming that the encoder respects your bitrate settings exactly, which it never does either with CBR or VBR. If encoded at comparable settings (vbr max does not exceed CBR level) X-pass vbr can achieve the same relative level of quality as CBR and do it in a smaller filesize, but it can never surpass it in quality. Again, this is all only referring to bitrate allocation.

Now the additional passes done during x-pass vbr can improve the motion search precision, which can increase quality. In this respect x-pass vbr theoretically can achieve better quality than CBR even though it never has more bitrate allocated to any given scene.

Realistically speaking, this added motion search precision isn't all that significant on most sources. For all intents and purposes, if using comparable bitrates CBR and x-pass vbr should look identical. You may, however, find the odd scene here and there where one mode or the other appears to achieve better quality.

The .vaf file is where all the info is stored that the encoder accumulates after each pass. It then uses this information to improve the bitrate allocation for each subsequant pass, updating the .vaf each time. If encoding in CBR it does not matter whether you check the .vaf creation option or not. You only need the .vaf file if you want to do multipass vbr or if you want to use the advanced settings in cce, but then thats basically x-pass vbr anyway.

iceman,

1. CBR can beat x-pass vbr. Whether it does or not depends on your settings and on the effectiveness of the encoder's multipass vbr encoding.

2. Temporalsmoother and any noise filter can have a dramatic effect on your output. Whether or not you see the difference depends on your settings and your eyes. In analyzing the effects of these filters look at fine details such as the lines on peoples skin or detailed objects in the background such as wallpaper or other textures. Too much of either filter can soften the image, removing these fine details. Without these filters the source becomes less compressible, and you run the risk of getting more artifacts in your encode. On clean sources these filters should be used sparingly, if at all.

3. Ok.

iceman
18th July 2002, 07:56
Thank you for your thoughts Adam. However I found your views conflicting with your own statement.

Your said

"if the CBR level is equal to your max bitrate in x-pass vbr than CBR can never be worse than x-pass vbr".

and

"Now the additional passes done during x-pass vbr can improve the motion search precision, which can increase quality"

That is my point, even though you use the same bitrate allocation still VBR comes on top. I am not an expert what I believe it is to do with it's ability with motion search precision (As you pointed out).

As far as Temporalsmoother and any noise filter are concerned they did not have a dramatic effect on my output. Mind you I might have to do more tests with different settings.

Your thoughts on .vaf file were very excellent, now I understand what it is!

Assuming it is good source and also the same bit rate(From which I am encoding), what other setting would you use to beat VBR (x passes)? I have tried deinterlacing as I said, it makes a big difference.

Thanking you.

adam
18th July 2002, 18:02
iceman when I stated that,

"if the CBR level is equal to your max bitrate in x-pass vbr than CBR can never be worse than x-pass vbr".

I also quantified that statement by saying,

"Again, this is all only referring to bitrate allocation."



The motion search precision is an entirely different factor. I broke the discussion up into parts and explained the benefits of each encoding mode according to both of these factors.


"Assuming it is good source and also the same bit rate(From which I am encoding), what other setting would you use to beat VBR (x passes)? I have tried deinterlacing as I said, it makes a big difference"

It's all relative. Whatever you do to improve the CBR encode you could also do to the VBR encode. As your overall quality increases the perceived quality differences between these two encoding modes will become less noticable, but its still there. It only seems like your shortening the gap between CBR and VBR, but your probably not.

iceman
19th July 2002, 08:16
Thanks for your thoughts. You know your stuff!

SiliconSoul
21st July 2002, 20:40
with using CBR i think u guys are creating an Xsvcd ( out of spec svcd). If Im thinking correctly the svcd and vcd spec call for vbr. im not sure if it matters although most players will play a cbr encode.... :-) just throwing out some ideas...:D

adam
22nd July 2002, 05:00
Actually the VCD specs call for CBR only and SVCD supports either CBR or VBR.

Labersack
22nd July 2002, 10:45
@Siliconsoul
you're wrong, with VCD only 1150kb CBR Video is allowed.
SVCD specs says VBR, but isn't CBR special version of VBR? VBR with max=min=avg?

SiliconSoul
23rd July 2002, 00:18
ahh i did not know vcd was supposed to be cbr... but ur right cbr is just vbr with the max and the min the same :-)

someguy_1999
28th July 2002, 06:37
I believe you have that backwards. VCD is definately CBR. SVCD, If i remember correctly, specifies CBR (constant Bit Rate), however I think there may be a provision for VBR (Variable Bit Rate) in the standard.

You may want to look through the guides/Faq's, because I'm Certain I read that here.

SiliconSoul
28th July 2002, 10:13
you are wrong... i can retype this but id rather not so here is the link to the svcd spec. and it SAYS "IT is based on variable bitrate mpeg2 encoding..." read it for yourself...

http://www.licensing.philips.com/includes/download.php?id=2698&filename=2450.pdf

could not copy and paste protected pdf

if that link does not work this is the one before it...

http://www.licensing.philips.com/information/cd/video/documents575.html

btw the link changed from the old one so u make me do some work :D :D :D :D

LOL! im just joking the link changed but i found it. :cool:

Mosaic
2nd August 2002, 21:27
To get the benefits of VBr u need to have a large peak and average max bitrate.

Also u need to have a LOW minimum bit rate .......and the AVERAGE minimum bitrate is the MOVIE overall average.

That way for hi, motion scenes the VBR will allocate lots of data ...like 7Mbits/s and for low colorless scenes...dark cloudy nite......it willl drop to the minimum.

Calculate movie min avg bitrate with size with: 585000/mins - audio kbps=video Mbps.

Don't go having the average bitrate outside the min max limits or u defeat the VBR benefits.

CBR is suitable if u can maintain 3500Mbs or better thru out .....less than that means go VBR. I generally use VBR 2 pass for 3pass with VAF. Don't use vaf with CBR .....no need.

someguy_1999
15th August 2002, 06:30
ok-ok-ok, I guess I got confused by tmpgenc, etc. (SVCD profile defaults to CBR.):o