On Mon, Aug 25, 2008 at 12:09:56PM +0200, Kevin Bortis wrote:
> Hi,
>
> Is there a way to encode DiracPro video under linux or mac os x at the
> current development stage? I have here some uncompressed 4:2:2 video
> files with Dolby A/52 audio I would like to convert to DiracPro
> embedded in mpegts files.
>
> I have read some ffmpeg docu and have compiled ffmpeg under linux with
> libschroedinger support. Do I have only to add the "-intra" option to
> get DiracPro instead of Dirac?
>
> If someone could give a working ffmpeg example I would be very happy.
This is what I use for intermediate work in HD720. Constant bitrate
at 100 Mbit/sec, disable arithmetic coding, intra-only:
gop-structure=1 enable-noarith=1 rate-control=1 bitrate=100000000
I use 100 Mbit/sec as a bit rate because it's nearly lossless and
disk usage isn't a concern for me. Another, but variable bit rate:
gop-structure=1 enable-noarith=1 rate-control=0 noise-threshold=50
I'm not certain how these convert into ffmpeg options. I use gstreamer
almost exclusively, which exposes the schroedinger options directly.
Explanation:
gop-structure=1: Intra-only. There isn't an explicit "use VC-2"
option currently, but this does exactly that. You pretty much
need this for any intermediate format currently.
enable-noarith=1: Uses VLC coding instead of arithmetic coding.
VLC coding is a *lot* faster for both encoding and decoding, which
is convenient at really high bit rates.
rate-control=1: Encodes at a constant bit rate. Each picture is
usually within a few percent of the target size.
bitrate=100000000: Bit rate, in bits per second.
rate-control=0: Encodes at a constant quantisation noise threshold.
The quantisation noise will never be higher than this threshold,
although it can often be lower. This is a variable bit rate
method that isn't very good.
noise-threshold: The quantisation noise threshold converts directly
into quantisation factors, and is roughly correlated with PSNR.
Optional:
au-distance=1: This adds a sequence header to every frame, which
is good for seeking.
Future directions:
I'm currently working on a rate distortion method for variable bit
rate which is much better than the current "constant noise threshold"
method. One of the points of using a variable bit rate method is to
spend bits wisely at a global level -- i.e., use bits to compress
pictures, not noise. CNT has the problem that it spends a lot of
bits on pictures that have noise levels above the noise threshold.
Also, at some point, I'd like someone to implement (or hire me to
implement) a GOP structure that is highly seekable, but uses some
motion compensation. The most relevant limit is "number of reference
pictures that need to be decoded to decode picture N". For intra-only,
this is 1. For a highly seekable stream, this should be small, say
2-4.
Both of these techniques would be useful for intermediate storage.
dave...
-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/
_______________________________________________
Schrodinger-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/schrodinger-devel
