[Ffmpeg-devel] Embedded preview vhook/voutfb.so /dev/fb
Fri Mar 30 07:50:40 CEST 2007
On 3/28/07, Michael Niedermayer <michaelni at gmx.at> wrote:
> for which application that is something i dont know, if i had the choice
> between a working well written dirac decoder and a working well written
> filter system then i certainly would choose the filter system but a
> working dirac decoder/encoder is better then a failed attempt to implement
> a filter system and a dirac decoder is just implementing whats in a
> specification while the filter system could turn out to be very tricky
> that is the combination of direct rendering, slices and frame order
> inserting and removing of frames and arbitrary filter graphs could turn
> out to be fairly complex ...
> i think that bobby should comment on what he thinks would be a better
> task for him (that is what he would prefer working on and if he thinks
> he can actually design and implement a system supperior to libmpcodecs
> that is without any filters ...)
The more I think about it, the more the filter system sounds more
interesting to work on. I plan to look at various systems out there,
and various ideas that have been thrown around, and take the best of
what I find, along with any new good ideas I might hit upon myself.
And since I'll be discussing the design here as it progresses, bad
ideas that creep in should get killed off pretty quickly, and I think
the result should end up at least as good as libmpcodecs.
Of course, while I certainly hope to do even better, there is the
possibility that the end result turns out very similar to libmpcodecs.
You've suggested in the past that people port it to ffmpeg, so I
think that'd still be a fair improvement, but a simple port wouldn't
be worth a whole SoC project.
I'd still be happy to work on Dirac, but reading Rich's ideas, and
starting to go back and look at libmpcodecs and the mplayer-g2
discussions, I'm much more excited about trying to create a better
More information about the ffmpeg-devel