CQ better than 2-pass VBR?

[SiCN]

Hi,

I am wondering which one is better. CQ (Constant Quality) or 2-pass VBR.

I have been encoding my SVCDs with a VBV buffer at 224 and a max bitrate of 2520 and a min bitrate of 0 (resolution: 352x576, PAL). Now Tmpgenc says that CQ will try to keep a constant quality over the whole movie and 2-pass VBR says it will first scan the movie and then individually set the bitrate according to the material.

Isn't it better to have it on CQ at like 90% with max bitrate of 2520 instead of setting 2-pass VBR which takes twice as long to encode? Why do you need to set average bitrate in 2-pass VBR if it is anyway supposed to scan and decide itself on the first pass?

Anybody have a good explenation?

Thanks!
/Daniel

[DaFoigna]

Well, I believe that 2pvbr is pretty much the same as cq. The procedure for both encoding methods is to analyze the video sequence for motion and then encode the stream so that high motion is given a higher bitrate than low motion. The quality should be the same throughout the video sequence.

But when you encode a video sequence using cq method you actually never know the size of the encoded file. I always use 2pvbr so that I can calculate the (almost) exact size of the video sequence so that there is almost no unused space on the CD.

Example: 2hr video sequence encoded in 2pvbr to fit on 2 80-min CDs. Set max. bitrate to 2376 kb/s, min. to 0 kb/s. Average bitrate can be calculated as follows: (2(no of CDs)*800MB(Space on 1 CD)*1024*8/60/120min(length of video sequence))-224(audio-stream)= appr. 1596 kb/s. I recommend deducting about 2-3% in video stream rate so that you don't have to cut the movie in the middle of a scene. Therefore, I use 1550kb/s as average video bitrate. Use 224 kb/s as audio bitrate. The file size should be appr. 1560MB which fits right onto 2 80-min CDs.

Yes, it takes a long time to run through the video sequence twice, but the quality is great, and the space usage on CDs is optimized. I do this when I encode SVCDs for best results. Try it...

Regards,
DaFoigna

[SiCN]

Okay! Thanks, but why does Tmpgenc need to know the "average bitrate" if it is supposed to decide itself on 2-pass VBR?

[Kdiddy]

He just told you why...

"But when you encode a video sequence using cq method you actually never know the size of the encoded file. I always use 2pvbr so that I can calculate the (almost) exact size of the video sequence so that there is almost no unused space on the CD. "

The average is your bitrate used for calcualtion purposes to know what fits on a disk..with CQ you do not have this..so you have no clue what the final size will be & whether or not it will on a disk.

[IGuaNAdoN]

if i don't really care if it fits on the cd or not... will CQ give better or the same QUALITY as 2passVBR. i have always used CQ and if it took 3 cds so be it... i have another question also... i use CQ 75 max 2850 min 2150. after mpg is encoded i open it in Bitrate Viewer 1.5.002 the bitrate is yellow line and Q. level is green line. it seems that the Q. Level has an effect on my DVD player... if the Q. Level jumps way up it seems my DVD player halts at this point. ex bitrate one frame before is 2284 Q. Level 2.53 next frame bitrate 2725 Q. Level 9.75. I have higher bitrates in the file but the Q. Level(Quality?) is not near as high as 9.75 ...highest is 2799 Q. Level 3.87. does this mean anything to anyone? maybe you could point me in the direction of how the quality slider bar affects playback or does it at all? Thanks!!!

[VidGuy]

My experience has been that CQ will give better quality than 2-pass VBR (where, for me, quality = # visible macroblocks) for the same size file.

Of course, the trick is - how do you get the same size file - I experimented for several days on this to determine which I liked better. YMMV.

[kinneera]

My understanding of CQ is that it sets the quantization level constant throughout the entire movie and then basically does 1-pass VBR on it. The problem with this is that a constant quantization means that the bitrate has to be sufficient to reach that level no matter the complexity of the scene. This means that if it is uncapped (true CQ), you can get humongous bitrate spikes. TMPGEnc appears to use a sort of "capped" CQ. Personally, I don't think there is enough of a quality difference (if any) to accept its unpredictability in filesize over the predictability of multi-pass VBR. The general guideline, though, is that movies with lots of motion are going to produce much larger files.

[KennyC]

<TABLE BORDER=0 ALIGN=CENTER WIDTH=85%><TR><TD><font size=-1>Quote:</font><HR size=1 color=black></TD></TR><TR><TD><FONT SIZE=-1><BLOCKQUOTE>
On 2001-12-19 12:17:40, kinneera wrote:
The general guideline, though, is that movies with lots of motion are going to produce much larger files.
</BLOCKQUOTE></FONT></TD></TR><TR><TD><HR size=1 color=black></TD></TR></TABLE>
Hi,

I'm glad someone thinks the same way as me. I'm working on various templates for Fast, Normal and Slow Motion movies at present using CQ 100% to see what differences if any the bitrate variations have on the size and quality of the movie.

Initial tests with The Mummy Returns with all it's warrior scenes which shows up macroblocks well is very encouraging.

I agree with others, the things I hate most of all are:
1. MacroBlocks
2. Scene pauses
3. Scene jumps

I would rather watch 3 CD's of SUPERIOR quality that 1 CD of INFERIOR.

I'm also trying to finalize complete UNLOCK and LOCK templates. (I know TMPGEnc already provides a kinda unlock.mcf)


©.¸¸.·´¯`·.¸¸.·´¯`·.¸¸.·´¯`·-- KennyC --·´¯`·.¸¸.·´¯`·.¸¸.·´¯`·.¸¸.©

[KennyC]

<TABLE BORDER=0 ALIGN=CENTER WIDTH=85%><TR><TD><font size=-1>Quote:</font><HR size=1 color=black></TD></TR><TR><TD><FONT SIZE=-1><BLOCKQUOTE>
On 2001-12-19 11:23:55, VidGuy wrote:
My experience has been that CQ will give better quality than 2-pass VBR (where, for me, quality = # visible macroblocks) for the same size file.

Of course, the trick is - how do you get the same size file - I experimented for several days on this to determine which I liked better. YMMV.</BLOCKQUOTE></FONT></TD></TR><TR><TD><HR size=1 color=black></TD></TR></TABLE>
Hehehe, I'm curious, which settings did you like best. Cheeky aren't I.