On Sun, 14 Nov 2004, sean wrote:
> Yes. I'm going to try another dv camcorder to see if that makes a
> difference. It may also be that I haven't played the copy on a tv, just
> on my computer monitor. Maybe the tv will smear things out?
TVsets do smear/blend the interlacing/combing effect but not the
"blocks" that you mentioned.
Hmmm, what are you using to play the files on the computer? MPlayer?
I wonder if (some) of the stuff you mentioned is playback/decoding
related.
Can't say I've heard of a camcorder goofing up the analog->DV conversion
before but I suppose anything's possible.
> > The first thing in the "chain" was a color-corrector/image-stabilizer.
> > Something like the SCC-2 from http://www.simacorp.com/ (at the time
> You mean before any software?
Exactly/precisely what I mean. You want to present as stable an image
as possible to the capture-card/conversion-device - and that's
obviously happening well before any software gets a chance to run or
manipulate the data.
A while back there was a thread about the MJPEG cards (DC10+ or DC30)
dropping frames (which leads to A/V sync issues later on) because
the tape was less than great quality and the VCR was presenting an
unsteady signal.
> I'd always thought that software like mjpegtools was just as "good" but
> not as fast as hardware.
Software (such as 'yuvcorrect') can do color correction after the
data has been captured but to regenerate the sync/colorburst info
from a tape has to be done outboard in the hardware before the
system gets the data.
> IOW, should I get the ssc-2 even if I'm going to use mjpegtools as you
> discuss below?
When I started out it was with a WinTV card and an earlier Sima SCC
model (long discontinued and replaced by the models you see now. I
look back on those videos and shudder but at the time it was the best
that could be done. Later the Canopus ADVC100 replaced the WinTV card
but the Sima unit was retained even though the Canopus units are quite
a bit more tolerant of unsteady signals - the image was noticeably
steadier with the correction unit inline and that equates to a higher
quality video (and/or a lower bitrate ;)).
Oh, as a side effect of regenerating the sync and colorburst info
any macrovision crud is stripped out (left behind actually ;)).
The macrovision signals can confuse capture cards in the same way
that the signals are meant to confuse other vcrs (that's how the
copy protect works...).
> > After that it was into a Canopus ADVC-100 (today that would be a
> > ADVC-300 which has hardware denoising capability as well as a TBC
> > (TimeBaseCorrector - stabilizer).
> >
> Same question. Now I'm just using the i1394 connection.
Today if I were starting over I'd get the ADVC300 and skip the Sima
unit since the -300 has a TBC builtin (and denoising circuitry). I
do not know (since I don't have any VHS tapes left) if the Canopus units
refuse to convert analog if macrovision signals are detected or not (the
manual says that the status light will blink).
> Yes. Bigger is better. I'm actually angling for an AMD64 machine next.
> What's your read between dual and 64?
As you say 'bigger is better' - which means a dual 64bit system of
course :)
All depends how much $$ you have. If you want a dual cpu system
on the cheap then a dual Athlon-MP barebones (cpu, memory, motherboard)
can be had for just a little more than a single cpu E-machine.
If you have a surplus of funds then a dual Opteron 250 system would
be the way to go.
Actually at this point in time and for some time to come (unless brave
volunteers step forward ;)) no matter WHAT system you get you'll want
to build and run 32 bit binaries for mjpegtools (and a fair amount of
other multimedia software). Why? Because the critical assembly
language routines that provide speed in key areas haven't been
rewritten/ported to 64 bits yet.
> I can see noise around the edges on the video. I assume I want to convert
> all that noise to black.What's the mjpegtools equivalent of -J mask in
> transcode. I masked 8 pixel on a side, which seems to get rid of that
> noise, and I assume will be in the overscan region anyhow.
Yes indeed it'll be in the overscan area which means viewers on a TV
won't see it - but the encoder will and will waste a lot of bits
encoding rubbish.
Many ways to black the border - BUT, if you simply want to crop the
8 pixels from each side (and 704x480 is the NTSC frame size not 720x480)
then y4mscaler will do that:
y4mscaler -O sar=src -O size=704x480
just feed in a 720x480 and get 704x480 out (and yes, 704x480 is a valid
framesize to put on DVD).
If you want to preserve the 720 frame size you can use 'y4mshift'
y4mshift -b 8,0,704,480
and if the video isn't exactly centered you can add "-n X" or "-N X"
to shift the video right/left or up/down. y4mshift -h will get the
usage summary that gives the info.
Either of the scaling programs can put borders on - for that I'll
refer you to the manpages or usage() statements for 'yuvscaler' and
'y4mscaler'.
And finally the 'yuvdenoise' program has a '-b' option that can black
out specified areas during the denoising process.
My preference usually is to crop to 704x480 - gives slightly smaller
files since there's less data to encode ;)
Cheers,
Steven Schultz
-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8
_______________________________________________
Mjpeg-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/mjpeg-users
