> From: Matto Marjanovic <[EMAIL PROTECTED]>
>
> How about trying it without yuvdenoise in the pipeline?
> yuvdenoise does quite a bit of low-pass filtering of its own, which means
> that later stages are less prone to aliasing ('cause it leaves less HF
Ok - here you are - and this almost certainly will be of interest
to those making SVCDs from DV (Digital8) data.
> >Does the smaller output size actually mean y4mscaler is "better"?
>
> Good question --- what's it look like, Steve? :)
Looks good!
The results below are addressing two questions at once (B vs P frame
size and yuvscaler/y4mscaler comparison).
~45 seconds of DV data (Digital8) encoded to SVCD and DVD.
Using "yuvscaler -M BICUBIC -O SVCD":
INFO: [mplex] No. Pictures : 1348
INFO: [mplex] No. Groups : 90
INFO: [mplex] No. I Frames : 90 avg. size 21052 bytes
INFO: [mplex] No. P Frames : 361 avg. size 7551 bytes
INFO: [mplex] No. B Frames : 898 avg. size 6420 bytes
INFO: [mplex] Average bit-rate : 1846000 bits/sec
INFO: [mplex] Peak bit-rate : 2360400 bits/sec
Using "y4mscaler -S option=cubic -O preset=svcd":
INFO: [mplex] No. Pictures : 1348
INFO: [mplex] No. Groups : 90
INFO: [mplex] No. I Frames : 90 avg. size 21295 bytes
INFO: [mplex] No. P Frames : 361 avg. size 7150 bytes
INFO: [mplex] No. B Frames : 898 avg. size 6084 bytes
INFO: [mplex] Average bit-rate : 1770400 bits/sec
INFO: [mplex] Peak bit-rate : 2330800 bits/sec
Looks like ~15% difference between B and P sizes at SVCD resolution.
It also appears that there's a ~4% bitrate reduction when using
y4mscaler to get SVCD sized frames.
And lastly no scaling - going direct to DVD:
INFO: [mplex] No. Pictures : 1348
INFO: [mplex] No. Groups : 90
INFO: [mplex] No. I Frames : 90 avg. size 29793 bytes
INFO: [mplex] No. P Frames : 361 avg. size 11778 bytes
INFO: [mplex] No. B Frames : 898 avg. size 10023 bytes
INFO: [mplex] Average bit-rate : 2832000 bits/sec
INFO: [mplex] Peak bit-rate : 4241200 bits/sec
at the DVD frame size the difference is ~10% between B and P frame
sizes.
I did use the "-N" to mpeg2enc but yuvdenoise was _not_ used in any
of the above tests.
Cheers,
Steven Schultz
[EMAIL PROTECTED]
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
Mjpeg-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/mjpeg-users
