Low Compression >80% DVDrebuilder Vs Shrink? [Archive]

casperse

25th April 2005, 20:43

Hi Guys

I wanted to make a small test to see the differens between encoding DVD movies with small compression 80-99% with DVDrebuilder and ex Shrink.

I am using Rebuilder v83 and Cinema Craft Encoder SP v2.67.00.27
4 Passes & VBR Bias= 25 & Quality_prec=24 (24 and not 16 because read LINK (http://forum.doom9.org/showthread.php?s=&threadid=74460))

In one of the other Post I read that some one said that:
>>DVDShrink with ACE enabled (Smooth) produces the same or better quality than Rebuilder with CCE at 80% compression and above"

I love DVDrebuilder and I always use it, this is just to see if the EASY fast encoding (1-2Hours) of movies with a really low compression level could be done the quick way without a big Picture loss vs DVDrebuilder?
And also to see if a encoding fast following the org bitrate would be better when so little encoding is nessesary?

So I decided to make a small test :D to see if I could see any differens.

******************************************************************
See pictures below, From "Sky Captain" Compression ONLY 86,9%
Frames 5742,6844,25810,33251,33575
All pictures is grabbed in BMP, and saved in Photoshop as JPG at maximum 12 level to keep them 99,99 to org!

ORG http://img128.echo.cx/img128/3270/5742org869pct6xv.th.jpg (http://img128.echo.cx/my.php?image=5742org869pct6xv.jpg)DVDREBUILDhttp://img120.echo.cx/img120/9351/5742dvdrebuildr869pct0sv.th.jpg (http://img120.echo.cx/my.php?image=5742dvdrebuildr869pct0sv.jpg) SHRINK http://img120.echo.cx/img120/32/5742shrink869pct6ym.th.jpg (http://img120.echo.cx/my.php?image=5742shrink869pct6ym.jpg)
ORG http://img128.echo.cx/img128/4621/6844org869pct1kk.th.jpg (http://img128.echo.cx/my.php?image=6844org869pct1kk.jpg)DVDREBUILDhttp://img120.echo.cx/img120/2508/6844dvdrebuildr869pct9en.th.jpg (http://img120.echo.cx/my.php?image=6844dvdrebuildr869pct9en.jpg)SHRINK http://img120.echo.cx/img120/6349/6844shrink869pct6xn.th.jpg (http://img120.echo.cx/my.php?image=6844shrink869pct6xn.jpg)

ORGhttp://img128.echo.cx/img128/4958/25810org869pct2fs.th.jpg (http://img128.echo.cx/my.php?image=25810org869pct2fs.jpg)DVDREBUILDhttp://img120.echo.cx/img120/511/25810dvdrebuildr869pct3er.th.jpg (http://img120.echo.cx/my.php?image=25810dvdrebuildr869pct3er.jpg)SHRINK http://img120.echo.cx/img120/3255/25810shrink869pct5ws.th.jpg (http://img120.echo.cx/my.php?image=25810shrink869pct5ws.jpg)

ORG http://img128.echo.cx/img128/4779/33251org869pct2mi.th.jpg (http://img128.echo.cx/my.php?image=33251org869pct2mi.jpg)DVDREBUILDhttp://img120.echo.cx/img120/9956/33251dvdrebuildr869pct8mq.th.jpg (http://img120.echo.cx/my.php?image=33251dvdrebuildr869pct8mq.jpg)SHRINK http://img120.echo.cx/img120/8728/33251shrink869pct4ty.th.jpg (http://img120.echo.cx/my.php?image=33251shrink869pct4ty.jpg)

DVDREBUILDhttp://img120.echo.cx/img120/3439/33575dvdrebuildr869pct8ra.th.jpg (http://img120.echo.cx/my.php?image=33575dvdrebuildr869pct8ra.jpg)SHRINK http://img120.echo.cx/img120/5144/33575shrink869pct3ik.th.jpg (http://img120.echo.cx/my.php?image=33575shrink869pct3ik.jpg)

Hmmm Looking forward to hearing what people think and what YOU do when you need to encode movies at only 80% and above compression :confused:
DVDrebuilder seems to be more light in some frames?
And Shrink is alittle sharper even if its encoded with ACE enabled and Smooth picture.

******************************************************************
Just noticed another Thing,On DVDrebuilder frames there is Noise in top of picture? See picture nr 2 very very small black spots like somebody nipped alittle of the picture?
I DID NOT used Crop function because there are no bars in this movie :confused: How can this be?
I used Deinterlaced with Decomb can this be the reason?
This is really a bad thing will have to test other movies done with DVDrebuilder :(

EDITED:
HMMM there is half black bar on the left top and on the bottom on the right, is this anamorphic ? If I Disable the Deinterlace with Decomb will that solve this?
******************************************************************

Fishman0919

25th April 2005, 21:21

Hummm, didn't I just say this about 30 min ago...oh well, here goes

Comparing frames to judge the quality of video is like comparing the taste of Ice Cream by looking at it... is just plain silly. The only real way of judging the quality of video is to watch it. One frame can't, can't, can't tell you how good the next frame is in relation to the one you are looking at... if you want to compare frame then JPEG might work well but not MPEG2.

If you want to compare between DVD-RB and DVDShrink maybe you can read this: http://forum.doom9.org/showthread.php?s=&threadid=91418&highlight=blind+test

is was with full disc compression but you could use it as a guide... further in the thread I talk about DVD-RB vs Shrink with Movie only and is was with 70% - 90% and the result were the same.

RaistlinMajere

25th April 2005, 21:25

these tests always fail to point out on thing.

the eye can't make out a slight bit of noise in one corner of one frame in 1/25 of a second. But you have the slightest bit of jerkiness in the motion of the assembled frames and you notice it a mile of. Correct me if I am wrong but don't transcoders cut the B frames which are vital in the smoothness of the video

Its a bit like the golden ear tests for audio. You can analyse it with signal to noise ratios all night long but the true test is in the usage.

Add the fact that re-encoding you can use filters such as de-interlace, resizing, picture cleaning.

onesoul

25th April 2005, 23:09

I used Deinterlaced with Decomb I recommend not using any optional settings if you are not absolutely sure what you are doing. For example this movie I highly doubt that it is interlaced, even if it is flagged interlaced. If that's the case I suggest you use the setting "disable interlaced" on the respective vts. That way the encoder will see it as progressive and dvd-rb will flag it as progressive.

Even if it was interlaced, I wouldn't recommend deinterlacing. And it sure did have a negative impact in your test.

I suggest you try the "disable interlaced" setting and following settings:
- gop 12 2
- cce standard matrix
- 2 pass
- vbr_bias=35
- quality_prec=16 for cce 2.67 and if you use cce 2.70 try quality_prec=25

If you can I'd suggest you use cce 2.70 although it is reported to be buggy due to some missing frames, so use it with care.

One last note:
If you had really a interlaced movie, I would strongly suggest using HC encoder (Edit: or Quenc), had much better results with it.

Cheers

jdobbs

26th April 2005, 03:06

Blah, blah, blah... more drivel.

Learn what you're talking about, or just don't do these ridiculous tests... the pictures are jpgs -- that speaks volumes in itself.

I'm outta here.

casperse

26th April 2005, 12:38

Sorry you feel that way jdobbs ;) This was not a attack on DVDrebuilder I love the Tool! :D

Thanks for the god feedback Im still a newbie when it comes to special settings and MPEG2, So this is a good way to learn!:D
And nice that I can talk to Experts like you!

Yes its JPG and no its not fair, would like to post 10G but thats not possible LOL. Just wanted to know what Others do when you need to encode movies at only 80% and above compression. What settings or alternative software you use etc.

I read somewhere that enabling Deinterlace (DCOMB) could never hurt...
And I have therefor almost always enabled it...But now it seems that was a big mistake...

Sticker:
*****************************************************************
Deinterlace with DECOMB
You must have installed the Decomb plugin and defined it's path in the DVD-RB SETUP window before you can use this option.
If Enabled, Interlace sources are deinterlaced with the fielddeinterlace() function of the Decomb Package. This is recommended if you plan on viewing your output primarily on a progressive display, such as your computer.
******************************************************************
Will this not mean that when you have a LCD/Plasma Screen it would be better to use it?

Onesoul: Thanks I will try that! Right away
Can I use Fit2disc to analyse the movie? (Interlaced)

onesoul

26th April 2005, 13:21

Use your eyes :). Load it in VirtualDub and check for comb lines especially on moving scenes, if you don't see that, it is progressive.

Btw, I think plasmas and Lcd's can handle interlaced content aswell.

Pal dvd's are more common to have the interlaced flag set wrong. I never heard of Ntsc sharing this "issue".

Don't forget that most of the time you don't need to mess with those interlaced or progressive settings, so check if they are right before changing anything.

jdobbs

26th April 2005, 13:31

Here are some important points:

1. Percentages mean nothing. 80% of 1000 is less than 70% of 1200.

2. Never compare individual pictures unless you are planning to watch the movie a frame at a time. Think of the old cartoons. If you have ever watched one you'd see how unnatural it looks when each picture is drawn pristine. Later computer animation purposely added distortion in movement to give a natural appearance. Pictures ain't video.

3. Never compare on easy-to-encode originals. Hell, even if one of the methods does a better job, they are both so good that no one can tell the difference.

4. Know how the reduction engines work. If you are familiar with within-the-compressed-domain reduction you know that on easy-to-encode originals the I-Frames or even the P-Frames could be left untouched. They are bit-for-bit matches to the original... but the B-FRAMES (usually about 80% of the output) can be trashed... and most "stops" occur on I-Frames.

5. Use a common sense approach before posting results. If you reach a conclusion that we've been wrong all these years and the sun comes up in the afternoon rather than morning... no one will listen to you. A quality multipass encoder does the most efficient job of allocating space and creating video. End-of-story, period, no argument. It is based upon mathmatics and physics. If you want to argue that point you might as well start talking astrology and psychic phenomena. Gee... maybe that's why they use it at the authoring houses, what do ya think? That doesn't mean the other choices are bad, though -- because for many people they are good enough.

6. Never use a single movie as a comparison. That's ludicrous.

There are a lot more. But I'm tired of typing.

In other words, it is very, very hard to prove a point without a good scientific approach and lots of data to use. Full motion comparison watching of uninterested parties who don't know which-is-which is the only way to tell quality.

onesoul

26th April 2005, 16:52

Originally posted by jdobbs
1. Percentages mean nothing. 80% of 1000 is less than 70% of 1200. I just add that Q factor is also important on compression and where you can see an encoder outperform a transcoder.
I've seen on same DVD with same overall reduction, a vts (3000kbps after reduction) having a Q factor reduced to 60 and other vts reduced to 30, so I bet this one would be a hard one to a transcoder.

Transcoding can look better on some sequences but you can't achieve balanced quality throughout the whole movie like in an encoder.

On some special cases I guess a whole transcoded film could be better than encoded one, but I'd put the chances one in a million, I don't know, I don't bother getting statistics right, why transcode when you can encode... ;)

dragongodz

26th April 2005, 17:21

On some special cases I guess a whole transcoded film could be better than encoded one, but I'd put the chances one in a million
far less than that actually. if you only buy the latest blockbusters packed with extras then thats one thing but there are many more older dvd release with little if any extras released aswell, at least here there is.

the fact is that it is a case by case situation though. know the type of dvd you are dealing with ,what is the original bitrate, how much are you reducing it, what is the likley effect of that much reduction with the different methods etc etc etc , and then you can decide what you wish to use.

think of each dvd like a snowflake, each is an individual. :D

But I'm tired of typing.
again and again and again ? ;)
maybe a sticky would save this sort of things popping up every few months ?

jdobbs

26th April 2005, 18:00

again and again and again ? I'm a glutton for punishment.

Rockas

26th April 2005, 21:22

Well... at least seems like casperse as taken the "highlights" from the "same frame" (it looks like that)... I've seen many "so called quality tests" in some "well known" sites made by some "experts" that didn't even had that "expertise". Do you want an example? here you have:
http://dvd.box.sk/articles8.php

When checking the Images... take a real close look :D

onesoul

26th April 2005, 21:56

Originally posted by Rockas
Well... at least seems like casperse as taken the "highlights" from the "same frame" (it looks like that)... I've seen many "so called quality tests" in some "well known" sites made by some "experts" that didn't even had that "expertise". Do you want an example? here you have:
http://dvd.box.sk/articles8.php

When checking the Images... take a real close look :D Didn't jdobbs do that test? :D:D:D Just kidding :)

Btw is the intervideo dvd copy as good as they say?

Rockas

26th April 2005, 22:15

Well... I love the Interface... love the Join 2 DVDs function but... I could never make it work :D ... It spits a DVD but it didn't work on any software players I have installed... even DVD Shrink couldn't open it... DVD ReMake is Cheaper :D

I never tested the 1 2 1 conversion... 'cause I simply felt I was been cheated by a guy that I don't know... I don't like that... I knew him I could always think that the test was a joke.

I'm not accusing Intervideo of anything (maybe... I don't know... just maybe - this is just a thought, a personal one - they should be a little more careful to whom they are paying commissions :D)... anyway... I use several software from them and I like it.

keep it UP

edit: anyway... as it has been discussed on this topic... the "Images" test is all wrong from the beginning.

casperse

28th April 2005, 13:39

Originally posted by onesoul
I recommend not using any optional settings if you are not absolutely sure what you are doing. For example this movie I highly doubt that it is interlaced, even if it is flagged interlaced. If that's the case I suggest you use the setting "disable interlaced" on the respective vts. That way the encoder will see it as progressive and dvd-rb will flag it as progressive.

Even if it was interlaced, I wouldn't recommend deinterlacing. And it sure did have a negative impact in your test.

I suggest you try the "disable interlaced" setting and following settings:
- gop 12 2
- cce standard matrix
- 2 pass
- vbr_bias=35
- quality_prec=16 for cce 2.67 and if you use cce 2.70 try quality_prec=25

If you can I'd suggest you use cce 2.70 although it is reported to be buggy due to some missing frames, so use it with care.

One last note:
If you had really a interlaced movie, I would strongly suggest using HC encoder (Edit: or Quenc), had much better results with it.

Cheers

I just finished with your settings and disabled Interlace.
But I still get the pixilation in the top of the picture? :confused:

And when you watch the movie you can see the encoded black line as black spots jumping up and down...

If you look at one of the ORG jpg pictures you can se the black line in the top of the picture and at the buttom, after encoding it in DVDrebuilder I get this pixilation any idea on how I can fix this?

I use the latest DVDrebuilder and the recommended Decomb etc.
And ENC 2.67.

Cheers

onesoul

28th April 2005, 15:01

Could you post a avs file of 1 cell of the movie along ecl settings?

Just a side note, when I recommended q_prec 16 with cce 2.67 and q_prec 25 with cce 2.70, I meant it when using cce separate of dvd-rb. At DVD-RB you should leave always at 16 since it does the necessary conversion between each version scale (0-64 to 0-100).

casperse

28th April 2005, 16:14

Originally posted by onesoul
[B]Could you post a avs file of 1 cell of the movie along ecl settings?

Great I have attached the ecl below.

And here is the avs file:
#------------------
# AVS File Created by DVD Rebuilder
# VOBID:01, CELLID:02
#------------------
LoadPlugin("C:\DVD-Tools\DVD Rebuilder\DVD Rebuilder v0-83\DGDecode.dll")
mpeg2source("N:\SKY CAPTAIN AND THE WORLD OF TOMORROW\D2VAVS\V01.D2V")
trim(8762,13792)

ConvertToYUY2()
AudioDub(BlankClip())

***************

Just a side note, when I recommended q_prec 16 with cce 2.67 and q_prec 25 with cce 2.70, I meant it when using cce separate of dvd-rb. At DVD-RB you should leave always at 16 since it does the necessary conversion between each version scale (0-64 to 0-100). [/B

Does this mean that no matter what version of CCE I use, setting the Q-prec to 16 in DVDrebuilder will use the best value?

I read in a link that 24 and not 16 was best in 2,67 becauseLINK (http://forum.doom9.org/showthread.php?s=&threadid=74460)), where it was 20-24 if I useed 2,67 compared to 2,50 ? :D

onesoul

28th April 2005, 16:52

LoadPlugin("C:\DVD-Tools\DVD Rebuilder\DVD Rebuilder v0-83\DGDecode.dll")
mpeg2source("N:\SKY CAPTAIN AND THE WORLD OF TOMORROW\D2VAVS\V01.D2V")
trim(8762,13792)

ConvertToYUY2()
AudioDub(BlankClip()) Was there anything in the blank line above? Don't cheat on me ;)

If not I don't know what say except for you to try another version of cce, 2.70 or 2.50 (2.70 preferable), or hc and quenc. And you don't need the audiodub(blankclip()) line except on cce 2.50.
Does this mean that no matter what version of CCE I use, setting the Q-prec to 16 in DVDrebuilder will use the best value? Yes, DVD-RB converts the value according the version of cce used.I read in a link I posted in my first post, where it was 20-24 if I useed 2,67 compared to 2,50 ? I use value of 16 in DVD-RB based on my findings, you could try testing it yourself and use what you find best. I suggest bias between 35-40 for example.

casperse

28th April 2005, 17:41

No I wouldnt Cheat you Onesoul :D

All the AVS files have a blank line there, I think it is because I sometime use the Filter to crop the black bars with RB-Opt maybee this gives a empty line when I dont use the feature?.

If I look at the Org Movie it looks like this:

***********************
-----------............

...........-----------
***********************

If the "*" is the picture then the "-" is a half black line on top left and a half black line bottom right, and when picture are encoded it causes a jagged black line in these two places
Very strange, in the beginning I thought of interlacing because its half a black line in top and bottom....

gonzalo76

28th April 2005, 19:31

Originally posted by onesoul
I suggest bias between 35-40 for example.

That's why I will never know the right settings here, if they really exist.

If you make a search about the right settings here, everybody has differents opinions.

You said V=35-40, doom9 guide says 30, afterdawn guide says 25, a french guide says 10.
Same thing happens with Q_Prec values, you said 16, doom9 says 20, afterdaw says 16, the french guide says Q_Prec=4 ?

So that's why I'll never know the correct settings here.
I know, I know... "it depends on the movie".
Does anybody really know for what are these settings and the correct values or not or everybody is guessing here?

EDIT: I forgot Fishman0919 in his blind test:
DVD-RB .74 - One Click, VBR Bias=0, Quality Prec=16, rest default.

Thank you all.

Fishman0919

28th April 2005, 19:47

There is NO correct settings, every movie is going to be different... no 1 setting is the see-all-end-all setting that is going to magically make every movie look like God himself encoded it....Thats why you....Do settings, Encode, Watch and Compare...repeat

Because what I, Joe, Tom, Dick, Harry, doom9 guide, afterdawn guide and a french guide says might not work in your world... everybodies going to have the "Best" settings but there is no way there going to work for everyone everytime.

Play, watch...find out what works for you. I know the settings I use work for me.

onesoul

28th April 2005, 19:52

@ casperse

Did you use RB-Opt for something?

Btw, don't crop any black bars unless you add "pure" black borders back again (addborders(..,..,..,..) or something). The encoder may have trouble with non standard size and your player can complain later. I suggest you don't mess with that.

@ gonzalo76

CCE default bias is 30, and I suppose this is the standard setting for DVD-9 creation or for a not so restricted bitrate encode while still inside dvd standard. Well since, we are fitting a DVD-9 size into a DVD-5, it makes mores sense to me to increase bias, and as I said on other posts, because it stabilizes bitrate which can provide balanced quality. A lower bias would favour high bitrate demanding scenes in detriment of some sequences specially scene transitions (bias=0 is 100% VBR, BIAS=100 is 100%CBR). And I tested it and had better results too. Let me say that the french guide is rubbish, quality_prec is suggested by cce manual to be in 16-40.

But I suggest you to try it out and see what fits you best.

Fishman0919

28th April 2005, 20:35

@onesoul

I think you might have your thinking of what vbr bias does backwards

form CCE SP 2.70 guide:

As V/C becomes smaller, a stream becomes more
like VBR keeping the quality of the image with heavily up and down in bitrate. As V/C becomes larger, a stream becomes even like CBR
keeping the bitrate constant but without stability of the image quality.

when bitrates are low, you need the flexibility of a low V/C bias so the encoder can allocate bits where they are needed and where not needed....as the bitrate get higher, a low V/C Bias it not as necessary because of the already high bitrate.

30 is the CCE default because it is a good, average, in the middle setting for avg bitrates (thats why CCE Basic has that setting)....

Anyway is all worthless if you don't play with the settings to see what looks good to you...

casperse

28th April 2005, 22:16

Originally posted by onesoul
[B]@ casperse

Did you use RB-Opt for something?

Btw, don't crop any black bars unless you add "pure" black borders back again (addborders(..,..,..,..) or something). The encoder may have trouble with non standard size and your player can complain later. I suggest you don't mess with that.]

No I dont use that on this movie since it has no black borders and I never resize! only crop and add black borders afterwards.

But in this case I am using default settings so it is till a puzzle?

Especially since its the org movie that have this half black line thats get pixilated after encoding?

onesoul

29th April 2005, 00:21

Let me put the whole quote from CCE SP 2.70 guide:V/C (VBR/CBR) : Fluctuation of the bit allocation:
In VBR streams, Cinema Craft Encoder SP performs bit allocation balancing the quality of image on its own valuation basis. The valuation basis can be changed by V/C value between 0 and 100. The initial value is 30. As V/C becomes smaller, a stream becomes more like VBR keeping the quality of the image with heavily up and down in bitrate. As V/C becomes larger, a stream becomes even like CBR keeping the bitrate constant but without stability of the image quality.
It is not applied to CBR stream.
When the distortion of complicated part is more noticeable than that of flat part, smaller V/C value may improve image quality of complicated part. On the other hand, when the distortion of flat part is more noticeable, larger V/C value may bring the better result. Although you can also adjust the setting for a certain part of the footage one by one at Local bitrate setting in VBR bit allocation (see page 55), V/C parameter can change the bit allocation through the footage more effectively at a time.

You didn't follow my thought, I think I explained it right, because when I said higher bias would increase stability I am referring to values such 35-40, it would still take advantage of the vbr encoding. And more, with lower bitrate available, the encoder which isn't perfect makes errors (they aren't exactly errors because the encoding is just doing as it was told, and that's why you have bias to tweak) distributing that same bitrate benefiting, with preference, high demanding scenes. As I said before, transition scenes like fading sequences which go from 700kbps to 7000kbps almost instantly, on that compression curve you can see visually the negative impact, to minimize that impact I increase bias but not drastically, and with better results.

To summarize it:
Lets take a DVD-9 and there you have for example avg 5000 kbps and max 9000kbps, now you reencode, and you'll have avg 3500kbps and max 9000kbps, now what do you want to do? give priority to complex high demanding scenes which would take precious bitrate to those flat and transition scenes which will suffer greatly or you want to balance the distributed bitrate to try get a more stabilized quality? I say increase bias to 40.

Fishman0919 you have done some nice tests, maybe you are willing to do some more, I suggest a full (progressive) DVD back-up with all defaults from DVD-RB and cce sp 2.70 with standard matrix. Then compare exactly the same but with bias 40. I'll accept whatever your results deliver (what could I do anyway? :)).

Ok, I promise not talk no more about this. You are all probably getting tired of me, I can understand that :).

Fishman0919

29th April 2005, 01:28

But increasing bias gives less stability, it tells you in the guide...if you want to give bitrate to those flat and transition scenes there is a setting for that....quality_prec

V/C Bias is just letting the encoder determine when a scene need more bitrate and when it doesn't and how much leeway it has...quality_prec put the bit to flat areas or edges depending on on the setting

One does one job the other does another, if you are get problems in your flat areas maybe want to uses a high quality_prec, it might help...but limiting the V/C for low bitrates is just putting a choke collar on the encoder.

Ex of V/C, if you set V/C to 0 in CCE SP 2.70 and encode a clip at a low bitrate (2000k) and then do the same clip with HC Encoder...you get almost identical bitrate distribution and both encoders look great...but if you do that same chip with CCE SP at V/C 30 (Default) CCE SP just can't keep up with HC, I've done it... tested it... shown it to people... and have them said so.

V/C 0 for CCE SP is letting the encoder do VBR just like every other encoder, full all out... but in some cases (higher bitrates) you don't need that.

onesoul

29th April 2005, 02:24

@ Fishman0919

IMHO you are misfollowing what cce guide says.One does one job the other does another, if you are get problems in your flat areas maybe want to uses a high quality_prec, it might help...but limiting the V/C for low bitrates is just putting a choke collar on the encoder. If there aren't sufficient bitrate available, quality_prec won't do any good.
Ex of V/C, if you set V/C to 0 in CCE SP 2.70 and encode a clip at a low bitrate (2000k) and then do the same clip with HC Encoder...you get almost identical bitrate distribution and both encoders look great...but if you do that same chip with CCE SP at V/C 30 (Default) CCE SP just can't keep up with HC, I've done it... tested it... shown it to people... and have them said so. One thing I agree HC performs likewise CCE with bias=0.
Now about the rest, I have my doubts...

I have tested a clip avg 3100 kbps and I tried every encoder version of cce with all kinds of settings bias, quality_prec and I tried HC. Well the encoder which gave better output was cce 2.70 with quality_prec=25 (in dvd-rb is 16) and Bias=40, I tried many combinations between Quality_prec and bias, these were the resulting values. I definitely didn't like cce at bias 25 even less at 0, neither HC, the most problems were specially, and I insist, on those fading scenes and objects against flat backgound. I even suggested to Hank315 to introduce a parameter like bias at which he told me that it was something he had already thought about. So I say when HC controls this parameter then it will be on par (or almost) with cce on progressive source. On interlaced source HC is already much better than cce and it isn't related to bias, something that I can't explain.

If you want to test the same clip I'll set up a link for you somewhere. Or maybe better I'll put up some samples encoded and you will see what I am talking about.

Fishman0919

29th April 2005, 02:42

We could go on and on and on... but the botton line is a higher V/C will not give you a "More stabile image", maybe a more stabile bitrate... it may give you an Image you like and that is pleasing to you and...yes it may fix the flat areas but at a sacrifice of over all Video quality. The two settings work in conjunction, one does one, the other an another.

I am not "misunderstanding cce guide says" As V/C becomes smaller, a stream becomes more
like VBR keeping the quality of the image not to much to misunderstand

If there aren't sufficient bitrate available, quality_prec won't do any good

Correct, so raising the V/C just limits the bitrate...choking the enocder and the result is poor video quilty at low bitrates

and I insist, on those fading scenes and objects against flat backgound

V/C is not the problem here but quality_prec...

Quantizer characteristics
This is a parameter for the balance of bit allocation between simple
and complicated part. The range you can set is 0 to 100.
As the value becomes closer to 0, a higher bit amount is allocated
to complicated part of the image. As the value becomes closer to 100,
a higher bit amount is allocated to flat part of the image. When the
value is close to 0, the mosquito noise at the edges (noise causing hazy
part along the edges, looking like flying mosquitoes) is less noticeable,
but the contouring noise (noise which looks like contour line patterns,
which appear in flat and wide areas, such as a dark background) is
more noticeable. The opposite occurs when the value is closer to 100.
The optimal setting depends on the footage, but roughly speaking,
16 to 40 is recommended. If the bitrate is relatively high, greater
value may bring better result.

again not to hard to misunderstand

But, I have said it earilier in this post...if it works for you...then they are the best settings.

onesoul

29th April 2005, 02:59

Correct, so raising the V/C just limits the bitrate...choking the enocder and the result is poor video quilty. It won't sacrifice video quality! Are you saying that dropping highest bitrate on some scene sequence from 8700 to 8500kbps will do any harm to the video quality? Those 200 kbps can make a lot of difference if you put it in a low needing bitrate sequence of 700kbps.

Bias is a powerful parameter which deserves to be embraced.

I will definitely put some samples up so I can least show my point. I'll do it tomorrow, I go to bed now.

edit: As I said before, when there isn't enough bitrate you can't tweak Quality_prec and expect it to do magic as it will show more mosquito or colour band or even both. Quality precision doesn't possess the blessing of bitrate multiplication. I have tried it!

onesoul

29th April 2005, 03:23

Correct, so raising the V/C just limits the bitrate...choking the enocder and the result is poor video quilty at low bitrates I saw you added the "at low bitrates" and I had to reply.
That's exactly the other way around!
Increasing bias will give better quality at low bitrates.
If you have all the bitrate to sell you can use bias=0 or OPV with Q=1 and you'll have the best encode possible unless you use cbr 9000kpbs.

Fishman0919

29th April 2005, 03:51

saw you added the "at low bitrates" and I had to reply. That's exactly the other way around!
Increasing bias will give better quality at low bitrates.
If you have all the bitrate to sell you can use bias=0 or OPV with Q=1 and you'll have the best encode possible unless you use cbr 9000kpbs

So CBR is better at low bitrates then VBR, by increasing bias you make the stream more like CBR... I'm more and more convinced your thinking on how V/C works is backwards... in your quote you contradict yourself... Increasing bias will give better quality at low bitrates... cbr 9000kbbs is like vbr v/c 100 and at a high bitrate... high bitrate, you don't need it.

If you got a lot of bitrate (4000 or up) V/C is not as need and can be raised.. as it said in the guide...lower bitrates movie need to be able to allocate bits better to preserve quality (raise and lower the bitrate when need as much as needed)to manage is bits better, a high v/c doesn't allow that... a movie at an avg bitrate of 2500 still can up up to 8500 or more...if you have the V/C turn up (35-40) you are limiting the encoder and it can't do it job right...low bitrates are where a lower v/c is need... high bitrate is not...as said in the guide.

onesoul

29th April 2005, 16:30

So CBR is better at low bitrates then VBR, by increasing bias you make the stream more like CBR... I'm more and more convinced your thinking on how V/C works is backwards... in your quote you contradict yourself... Increasing bias will give better quality at low bitrates... cbr 9000kbbs is like vbr v/c 100 and at a high bitrate... high bitrate, you don't need it. I am not contracditing myself, please don't distort my words, if you had tried to understand what I have been saying, you wouldn't see any contradiction, but you just keep promptly refuting without imo even thinking it through or even do some testing.If you got a lot of bitrate (4000 or up) V/C is not as need and can be raised.. as it said in the guide... The CCE guide doesn't say that!lower bitrates movie need to be able to allocate bits better to preserve quality (raise and lower the bitrate when need as much as needed)to manage is bits better, a high v/c doesn't allow that... And when those bitrates are raised and lowered, damage is done to scenes at low bitrate (and I am not going to specify again what those scenes are), increasing bias compensates them at practically no cost to high demanding scenes providing bias increased in a wise way.
a movie at an avg bitrate of 2500 still can up up to 8500 or more...if you have the V/C turn up (35-40) you are limiting the encoder and it can't do it job right...low bitrates are where a lower v/c is need... high bitrate is not...as said in the guide. The CCE guide doesn't say that. And I quote what I said before:It won't sacrifice video quality! Are you saying that dropping highest bitrate on some scene sequence from 8700 to 8500kbps will do any harm to the video quality? Those 200 kbps can make a lot of difference if you put it in a low needing bitrate sequence of 700kbps.
I will put the samples up when I get home, for many seeing is believing.

Fishman0919

29th April 2005, 18:17

but you just keep promptly refuting without imo even thinking it through or even do some testing.

Yes I have done test as I said before, the test was to see what settings worked the best for CCE SP at certain bitrates so I could use those setting in my encoder test... the results were that CCE SP worked better at bitrate between 2000-3000k with a low V/C of 15-0 and a low quality_prec of 16-12... 10 for v/c was a good middle of the road at those bitrates... at about 4000k raising the v/c to 20-30 and quality_prec to 30-35 help with image quality. Over 50 people viewed the clip and told me what they saw. CCE SP worked better with a low v/c at low bitrates. Period, end of story.

If you are going to refuse to listen and do what you think the guide is saying then so be it... if your settings work and looks good for you then great.

casperse

2nd May 2005, 20:10

Hi Onesoul

DVD back-up with all defaults from DVD-RB and cce sp 2.70 with standard matrix. Then compare exactly the same but with bias 40.

I am using Cinema Craft Encoder SP v2.67.00.27 would these settings be the same.

Quality_prec = 16
And Bias = 40

I would like to test this I had issues on the movie "Chicago" where there was steam/foog and it showed up very pixilated.

onesoul

2nd May 2005, 20:16

Yes casperse, with dvd-rb they are the same. But as I said before, if you can use cce 2.70 instead, it delivers (even) better results.

onesoul

2nd May 2005, 21:52

As I've promised, here comes the test clip encoded using bias 25 and bias 40 (V/C (VBR/CBR) : Fluctuation of the bit
allocation).

The test consists on the clip "for the birds" with lenght of 3:15. I decided to put up the whole encoded clip to be a more fair comparison.

Settings used for both encodes with CCE 2.70 (for the time being I consider it the best on encoding progressive source):
- 2 pass encode including creation of vaf file;
- 300min 3100avg 9000max (kbps);
- quality precision: 25 (at DVD-RB it is 16);
- dc precision: 9;
- progressive frame (although it was original wrongly flagged as interlaced), zigzag scan block scan order;
- GOP 12 2;
- cce standard matrix;
- adaptive q-matrix: off (actually i like it better off);
- no cce internal filter is used (they are not good!);

Script used:
LoadPlugin("C:\avisynth\neuron2\dg\DGDecode.dll")
mpeg2source("c:\dvd isos\monsters_inc\birds.D2V",idct=7)
fluxsmoothst(3,3)
converttoyuy2()
Encoded one clip with bias 40 and the other with bias 25 being each each one compressed in 3 parts as:
birds_bias25_qoff.part1.rar, etc.
birds_bias40_qoff.part1.rar, etc.

These files are hosted with kind permission of Mr. Donald Graft at (in onesoul directory):
ftp.neuron2.net
user: guest@neuron2.net (enter it just like that)
pwd: guest

I can give some pointers to find differences (which are noticeable:
- The lamp jumping around along the letters have much less mosquito with the bias40 encode.
- (Still in the first sequence) the background is less blocky with bias40.
- The fading is less blocky with bias40.
- Scene transitions with bias40 don't present as many as artefacts (such as blocks) as bias25.
- Clouds look better, less blocky on bias40.

I can say from thorough analysis that if high demandig bitrate scenes suffer with bias40, they are minimal. Overall quality is better with bias 40 and has a more constant quality to the eyes.

And here are a interesting analysis by bitrate viewer:
original:_____________bias40:______________bias25:_____________
http://img223.echo.cx/img223/1808/originalinfo7mk.th.jpg (http://img223.echo.cx/my.php?image=originalinfo7mk.jpg) http://img223.echo.cx/img223/1788/bias40info9nm.th.jpg (http://img223.echo.cx/my.php?image=bias40info9nm.jpg) http://img223.echo.cx/img223/5797/bias25info5jx.th.jpg (http://img223.echo.cx/my.php?image=bias25info5jx.jpg)

As you can see the peak bitrate for both bias25 and bias40 is respectively 7181 and 7220. Bias40 can actually achieve a higher bitrate. As you can easily see, the amplitude of bitrate oscillation of bias40 is smaller in comparison to bias25 which is not a bad thing.

Now I get to the more interesting part, Q level for the bias25 encode is more constant being it around 6. Remember what cce guide says that reducing bias (V/C) value you get more constant quality? Well it's exactly what I describe here, Q level gets more constant as bias approach 0 as the way a OPV encode would by selecting a determined Q.

Looking to the Q level of bias40 encode, it tries to attain same proportions as it was on original one. Which would inevitable lead (as verified in the visioning of the encoded clip) to a better and more constant quality, specially on those lower bitrate segments.

Conclusion from what is shown here:
In general bias(v/c)=40 won't prejudice any encode and probably will help them, specially on transition scenes, fading and low bitrate segments.

edited: to fix imageshack broken links

casperse

2nd May 2005, 23:12

Originally posted by onesoul
Yes casperse, with dvd-rb they are the same. But as I said before, if you can use cce 2.70 instead, it delivers (even) better results.

I would but EclCCE 1.81 dosent support the new version!
At least not on the forum page.

That is why I use the "Old" version

jdobbs

2nd May 2005, 23:23

???? You don't need ECLCCE with the new version.

erdoke

3rd May 2005, 00:22

As jdobbs suggested, using common sense approach.:D
Facts:
- Average bitrate will decrease on reencoding (when it performed to fit a movie onto a DVD5)
- Movies are different regarding optimal bitrate allocation.
- From the RB Settings section: VBR_Bias This setting controls how CCE distributes the bitrate. To quote jdobbs: "When you do your first pass the bit allocation per frame/gop is set based on a constant quality. The bias determines how much weight is given toward keeping that allocation. Setting it higher prevents wild swings in bit allocation between frames/gops. Setting it lower makes the encoder allocate more bits to high demand areas as needed and less to low demand. ...The higher you set this value, the closer you get to CBR rather than VBR."

My common sense approach tells me that using higher or lower than default VBR Bias value depends on:
- Source. Obviously if you have a movie with frequent dynamic scene changes (I mean from bright and rapidly moving to dark and almost still), it seems to be a better idea to let the encoder distribute more bitrate ( even from the lower average) to the eventful scenes from the dark scenes. And vice versa when your movie doesn't have many shifts of this kind.
- Source handling. Most probably there is a difference when you treat a movie as one "bit pool" and when you split it to smaller parts, just like Rebuilder does using the VOBIDs approach. I guess when CCE can work through the entire movie it will find more places to steal bitrate from in favour of the most bitrate hungry scenes.
- Resulting bitrate. It also seems obvious that if you have a resulting bitrate of 4500 kbps after a re-encode it will be different from an average of 3000.

I'm pretty sure that my common sense missed some arguments, but anyway it was not a scientific train of thought nor did I analyze movie parts encoded with different VBR Bias settings.;)

InFection

3rd May 2005, 14:02

well sorry but after reading the thread im totally lost! :(

in the RB help file it says :
For example:

Movies < 2hr:

quality_prec = 15
vbr_bias = 20

Movies between 2h - 2h10m:

quality_prec = 15
vbr_bias = 19

and so on..."

but now im reading numbers like Q_Prec : 25, 30 etc and for vbr_bias:30 to 40

is the help file outdated or.... ?

could somone recommend me good settings with the fallowing config:

DVD RB .88 pro
CCE SP Trail 2.70.02
movie aprox. 1h45min + extras (5.1GB)

i know the best is to test and find out for myself but since some of you already know the answer it will save me time...

Rockas

3rd May 2005, 14:54

Originally posted by InFection
well sorry but after reading the thread im totally lost! :(

in the RB help file it says :

but now im reading numbers like Q_Prec : 25, 30 etc and for vbr_bias:30 to 40

is the help file outdated or.... ?

could somone recommend me good settings with the fallowing config:

DVD RB .88 pro
CCE SP Trail 2.70.02
movie aprox. 1h45min + extras (5.1GB)

i know the best is to test and find out for myself but since some of you already know the answer it will save me time...

That's on the "Tips and Trick" section, isn't it.
Please read at the bottom of that page... those were tips from users of this forum... you can test them but as I stated on the help files:
"I didn't tested all of this tips personally."

edit: What I mean is... those tips can't be taken as absolute truths :)
That's why they are between " and are in Italic :D

jdobbs

3rd May 2005, 14:57

That was in the help file? I don't think I'd noticed that. I'll ask Rockas about it.

My recommendation is that it is best to leave the settings at the default for ALL sources. These settings (as the menu says) are for experts in CCE and are meant for specific purposes. I've tried to make DVD-RB tailorable for every level of expertise -- but I sometimes truly regret making some of these options visible. Most people change them and do more harm than good. Even worse, they change them for a specific DVD -- and forget them, only to wonder later why their output doesn't look as good as it used to.

Rockas

3rd May 2005, 15:02

Well... we posted at the same time :)

Hope my previous post makes everything clear :D

I guess I have to make all the warnings on a bigger font size ::(

InFection

3rd May 2005, 15:54

yep its in the tips/tricks section :)

ok i get it ..normal aka non experts should leave it to default BUT even for the Experts there must be some rule/logic so that they can chouse optimal settings depending on movie size and whether movie has alot of scene changes (dark /bright / etc) .

it would be great if that info can be posted and explained .

PS: @ Rockas & jdobbs ...i had posted before about the paypal button not working with some people (like me) ..remember ?

well i posted about it @ the unoffical paypal DEVS forum (whitch is normally not meant for user support) and they said to fix it you would have to do this :

You can either remove the '¤cy_code=USD' variable or change it to currency_code with the & in front of it.

Stephen Ivaskevicius
PayPal

hope it helps in some way

jdobbs

3rd May 2005, 16:06

It's kind of a contradiction, don't you think? Saying it's for experts and then posting rules. Understanding the impact of those settings is a lot more than a couple of rules. Saying "use X at this bitrate and Y at this one" would be wrong most of the time. If you know what it is doing internally (as an expert would) you can make your own judgements.

I say again. Leave them at the defaults.

jdobbs

3rd May 2005, 16:10

Originally posted by InFection
yep its in the tips/tricks section :)

ok i get it ..normal aka non experts should leave it to default BUT even for the Experts there must be some rule/logic so that they can chouse optimal settings depending on movie size and whether movie has alot of scene changes (dark /bright / etc) .

it would be great if that info can be posted and explained .

PS: @ Rockas & jdobbs ...i had posted before about the paypal button not working with some people (like me) ..remember ?

well i posted about it @ the unoffical paypal DEVS forum (whitch is normally not meant for user support) and they said to fix it you would have to do this :

hope it helps in some way Don't know what is happening to distort it in your case, but there IS a "&" there just before "currency_code". Unfortunately if you try to paste it here on the forum it gets changed...

onesoul

3rd May 2005, 16:39

@ erdoke

You have made valid observations but, Source. Obviously if you have a movie with frequent dynamic scene changes (I mean from bright and rapidly moving to dark and almost still), it seems to be a better idea to let the encoder distribute more bitrate ( even from the lower average) to the eventful scenes from the dark scenes. That's the opposite I have been showing/suggesting all this time, in this case higher bias would deliver better results.
For what I have seen, CCE works this way:
as bias gets lower the more constant Q level your encode will get and as you can see from my previous post, it doesn't mean constant quality (visually proven) because the original Q level isn't constant.

Considering it all, in my humble opinion a bias between 35-40 is a better default.

Rockas

3rd May 2005, 21:35

i had posted before about the paypal button not working with some people (like me) ..remember ?

Are you talking about the Button on Rebuilder or about the one on the site? (I had a little error on the site but that's fixed).

InFection

3rd May 2005, 23:25

yep just tried it now and works as it should ;)

Edsel

4th May 2005, 02:16

Originally posted by jdobbs
I say again. Leave them at the defaults.

And the defaults would be... ?

I've got Bias 25, Quality 24, but I think I changed them on advice of some posters here before.

SpazzHH

4th May 2005, 02:35

If I am not mistaken, the suggested Prec_Q should be set to 16. Some CCE versions use a scale of 0-64, others 0-100. Rebuilder automatically adjusts the number in CCE according to your version, so you don't need to make the adjustment yourself.

casperse

8th May 2005, 11:17

Originally posted by onesoul
I recommend not using any optional settings if you are not absolutely sure what you are doing. For example this movie I highly doubt that it is interlaced, even if it is flagged interlaced. If that's the case I suggest you use the setting "disable interlaced" on the respective vts. That way the encoder will see it as progressive and dvd-rb will flag it as progressive.

Even if it was interlaced, I wouldn't recommend deinterlacing. And it sure did have a negative impact in your test.

I suggest you try the "disable interlaced" setting and following settings:
- gop 12 2
- cce standard matrix
- 2 pass
- vbr_bias=35
- quality_prec=16 for cce 2.67 and if you use cce 2.70 try quality_prec=25

If you can I'd suggest you use cce 2.70 although it is reported to be buggy due to some missing frames, so use it with care.

One last note:
If you had really a interlaced movie, I would strongly suggest using HC encoder (Edit: or Quenc), had much better results with it.

Cheers

Hi onesoul

I would really like to get your input on settings for "Alexander" a movie that blows my settings (Yes I know you can only do so much when the bitstream is limited)

This is what I have after Editing:
Menu = 45MB
Movie = 6413MB (This is including 512MB 5.1 Sound + 12MB subs)
So I have a Approx 4465MB - 512MB -12MB -45MB menu = Approx 3895MB Left

This means I have 5889MB Movie that have to fit 3895MB.

I have cropped with settings Overscan 3, to squeze as much to picture quality as possible. (NO RESIZING! Aspect Ratio Error = 0%)
****************************
Crop(24,80,672,416)
AddBorders(24,80,24,80)
****************************

Settings: (Movie: Alexander)
CCE SP 2.7 (I have tried it an as you said it is excellent!)
QP = 16
VBR Bias = 40 (Tried it and it looked much better than when I did it with 25!)
Pass = 4
DC Presicion = 9
Q Matricies = Standard
And I have used RB-OPT to Compress The End TXT to a absolute maximum!
Giving more space to the actual movie.

(Im still a Newbie here and dont know if this Q Matricies should be altered?)

Any suggestions ? :D

jdobbs

8th May 2005, 13:20

Originally posted by Edsel
And the defaults would be... ?

I've got Bias 25, Quality 24, but I think I changed them on advice of some posters here before. If you go into your REBUILDER.INI file and delete those lines, the defaults will be used.

onesoul

8th May 2005, 14:16

@ casperse

Your settings seem fine to me and as long you like the results that's cool :cool:.

Personally I wouldn't do any cropping because I like to keep every video information, imagine one day you will be seeing the movie using a projector :) and most important, you are cropping precious vertical resolution (unless we were talking about a 2.35:1 movie but I wouldn't crop anyway)!

I really like cce standard matrix, I use it with everything, QLB is very similar and I use it when avg bitrate of the feature is around 2000-2500.

The last suggestion I give to you is that 4 pass isn't really necessary, generally I go with a 2 pass maybe 3 sometimes if you feel it's a hard movie but then again if you can afford the time let it pass away, harm it won't do :).

In the end it's all subjective and you should use what you find best.

casperse

8th May 2005, 20:52

Originally posted by onesoul
Your settings seem fine to me and as long you like the results that's cool :cool:.

I Do! :D
I was thinking about changeing the VBR Bias = 40 to 45 to see if it changes something, could see the differens with 25/40! Nice

Personally I wouldn't do any cropping because I like to keep every video information, imagine one day you will be seeing the movie using a projector :) and most important, you are cropping precious vertical resolution (unless we were talking about a 2.35:1 movie but I wouldn't crop anyway)!

Hmm If I only Crop the ORG and keep Overscan at 1 then it would be the same, only fifferent would be that the Black bars would be Macro Block "optimized" or did I get this wrong? That way I dont scrue with the aspect right?

I really like cce standard matrix, I use it with everything, QLB is very similar and I use it when avg bitrate of the feature is around 2000-2500.[/B]

QLB? :D What about DC Presicion = 9 or as default 10? & GOP=12

The last suggestion I give to you is that 4 pass isn't really necessary, generally I go with a 2 pass maybe 3 sometimes if you feel it's a hard movie but then again if you can afford the time let it pass away, harm it won't do :).

I only do this because I sometime get oversized result when I use passes lower than 4 :D and have size set to = CCETargetSectors=2260000 (A sidenote CC 2.7 Oversizes alittle more so I now have it at CCETargetSectors=2250000 and am testing it right now, didnt have any problems with 2.67 I guess something is changed)

onesoul

8th May 2005, 21:49

Originally posted by casperse
I Do! :D
I was thinking about changeing the VBR Bias = 40 to 45 to see if it changes something, could see the differens with 25/40! Nice Between bias 40 and 45 you won't be able to see much difference but if you feel that some sequences (at low bitrate, scene transitions and fading in/out) could be improved I suggest you test that clip using each setting.

Hmm If I only Crop the ORG and keep Overscan at 1 then it would be the same, only fifferent would be that the Black bars would be Macro Block "optimized" or did I get this wrong? That way I dont scrue with the aspect right? I bet you are using fitcd or fitdisk, right? :) Which is fine and will do a good job but in my opinion it isn't worth to mess wih cropping.

QLB? :D What about DC Presicion = 9 or as default 10? & GOP=12 In my tests I opted for choosing a dc precision of 9 which is a good value for medium bitrate and a good compromise between high motion/low motion features (dc 10 would best for low motion and dc 8 for high motion). Stick with dc 9.

QLB is Quenc Lower Bitrate which is a tweaked matrix by dragongodz, you can find it here (http://forum.doom9.org/showthread.php?s=&threadid=91821&highlight=qlb).
For using QLB in DVD-RB you can use Rocka's Rebuilder's Matrix Editor, you can find it here (http://forum.doom9.org/showthread.php?s=&postid=629239&highlight=quenc+lower+bitrate#post629239).

As for GOP, in my experience I find 12 2 (in cce you would choose M=3 and N/M=4) the best default, and as recommended I would only use GOP 15 2 (M=3, N/M=5) with NTSC (29.97 fps) pure interlaced material.

I only do this because I sometime get oversized result when I use passes lower than 4 :D and have size set to = CCETargetSectors=2260000 (A sidenote CC 2.7 Oversizes alittle more so I now have it at CCETargetSectors=2250000 and am testing it right now, didnt have any problems with 2.67 I guess something is changed) [/B] That's weird, 2 pass should suffice in achieving desired space target, I have used both hc and cce 2.70 (2 pass) with targetsectors=2264000 and obtained a 4480mb dvd ready to burn.