On Thu, May 06, 2010 at 04:45:42AM -0700, cullam Bruce-Lockhart wrote:
> 
> Hi there. Sorry if this is a repeat, I sent the same message out the other 
> day, but before I'd subscribed, so I think it got filtered out. Anyway, I'm 
> looking into the
> feasibility of using Schrodinger tools in a piece of
> software on board an Autonomous Underwater Vehicle. I've got
> an analog video camera on board the vehicle, which is
> converted to a digital YCrCb 4:2:2 feed over Ethernet through a Pleora IP 
> Engine. This video stream has a 720X520
> resolution, but is completely uncompressed. 
>  
> My big question, is how fast can the algorithms in
> Schrodinger handle medium quality compression? I'm assuming
> that intraframe compression would be faster than interframe,
> and I'd prefer for everything to be viewable as a still
> image anyway. So I guess I'd need it to do something along
> the lines of 720X520 intraframe compression at around 10-15
> frames per second, with a 10 hour video file being under
> around 40 GBytes, while running on a PC-104 (intel Atom
> processor) with a non-GUI Ubuntu operating system. Does
> anybody know if real time encoding at that level, on such a
> computer might be possible? 
> Thanks a lot for the assistance! 

As a slightly arbitrary data point, 4:2:2 720x520 @ 10fps,
vc2_simple or vc2_main profile, encodes in one core of a Core2
clocked at 1.67 Ghz at approximately twice real time.  Converting
to 4:2:0 first makes things faster.  Main profile Dirac (not
intra) can be configured at a variety of speeds and can be nearly
as fast as intra encoding, however, you would need to know
which undocumented magic parameters have an effect on speed.
It is a goal for a future release to expose this as a generic
"speed" setting.



dave...


------------------------------------------------------------------------------

_______________________________________________
Schrodinger-devel mailing list
Schrodinger-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/schrodinger-devel

Reply via email to