Downconverting from HDTV

[MarkGrigsby]

I'm downconverting some HDTV grabs using CCE SP, and getting some unexpected results... The grabs are either IVTC'ed 720p or 1080i.

I'm using a script similar to the one below to input to CCE, and the only video settings I change from default are to 1 pass VBR, 24fps, max bitrate 9800, M=1 and N/M=4. However, the output file is usually at about 3.4 Mbps avg, which seems far too low for high quality picture? Any clues where I'm going wrong?

loadplugin("C:\Program Files\dgmpgdec\DGDecode.dll")
mpeg2source("D:\DVD\Editing\Revelations\s01e1.d2v")
Video=last

Telecide(order=1, guide=1, post=2, vthresh=25)
Decimate(cycle=5)

LanczosResize(720,480)

FYI I use a similar method, with 'SelectEven()' to input true 60fps 720p material and the output is much higher bitrate with those same settings.

[Video Dude]

Originally posted by MarkGrigsby
I'm using a script similar to the one below to input to CCE, and the only video settings I change from default are to 1 pass VBR, 24fps, max bitrate 9800, M=1 and N/M=4.

Lower the Q value. I think the default is 60. Try between 20 and 40.

[MarkGrigsby]

Thanks - I did that and the bitrate shot up! But I don't understand why the results from the 30 progressive frame material are so different to the 24 frame stuff? The original quality of both is excellent, so how come the Quantization needs to be so different to achieve high quality outputs??!

I know there's never a 'correct' answer, but what do people tend to use for Quantization when dealing with high quality film-based stuff?

[Video Dude]

Low Q = high quality, big file size
High Q = low quality, small file size

Choosing a Q depends on the file size you want to end up with.

If you need help picking a Q value, use this program:
http://forum.doom9.org/showthread.php?s=&threadid=87764

[MarkGrigsby]

Originally posted by Video Dude
Low Q = high quality, big file size
High Q = low quality, small file size

Choosing a Q depends on the file size you want to end up with.


Thank you - I understand that, but what I don't understand is this... I feed a high quality 42 minute 24fps clip into CCE, and for a given 'Q' value the output is 720Mb. I then feed a high quality 42 minute 30fps clip into CCE, and for the SAME 'Q' value the output is 1.8Gb! That's what I can't figure out...

[Guest]

One is more compressible than the other. :)

[Video Dude]

Originally posted by MarkGrigsby
I feed a high quality 42 minute 24fps clip into CCE, and for a given 'Q' value the output is 720Mb. I then feed a high quality 42 minute 30fps clip into CCE, and for the SAME 'Q' value the output is 1.8Gb! That's what I can't figure out...


For one thing, a 30fps video has 6 more frames per second than a 24fps clip. That is about a 20% difference. So of course the 30fps will be larger if you are using 1 pass vbr.

Let's do some math:
6 additional frames * (42 minutes * 60)
= 15,120 additional frames for the 30fps video compared to the 24fps clip

24fps Video:
24 * (42 * 60) = 60,480 total frames

30fps Video:
30 * (42 * 60) = 75,600 total frames



Second,
24fps should be encoded at M=1 and N/M=4
but...
30fps should be encoded at M=1 and N/M=5



And what Donald said is true.

[MarkGrigsby]

Thanks (again!). I get 'Q values' now, but please would you just help me understand why 30fps should be encoded at N/M=5 instead of 4?

Cheers!

[Video Dude]

The formula is N / M = __

M is 3. (Don't ask why M=3 :D )


A frame is made up of two fields. Divide by 2.


Take 24fps. This has 12 fields. ( 24 / 2 = 12)
So, N=12, M=3,
12 / 3 = 4

Take 30fps. This has 15 fields. ( 30 / 2 = 15)
So, N=15, M=3,
15 / 3 = 5


Video compression is all math.

[danpos]

Originally posted by Video Dude

Video compression is all math. [/B]

I do agree with you!:D Ah, I suggest ccefront by Tylo because it uses RoBa method for achieve the Q value that 'match' the target filesize ...;)

Cheers,

[kolak]

Try 2 pass vbr encoding and then you'll know the final file size:)
CCE shows it to you.