Mailing List archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[linux-dvb] Re: Generating progressive video



On Tue, Feb 17, 2004 at 09:42:03PM +0100, Michael Plagge wrote:
> On Tue, 2004-02-17 at 21:23, Michael Hunold wrote:
> > On 02/17/04 21:05, Michael Plagge wrote:
> > > On Tue, 2004-02-17 at 20:53, Michael Hunold wrote:
> > >>On 02/17/04 15:33, Michael Plagge wrote:
> > >>>On Tue, 2004-02-17 at 14:35, Michael Hunold wrote:
> > >>>>On 02/16/04 23:38, Michael Plagge wrote:
> > 
> > >>>>Anyway, what solution (ie. which gfx adapter) do you have in mind to put 
> > >>>>the result to your beamer via DVI?
> > >>
> > >>
> > >>Have you checked if the DVI output is actually supported?
> > >>
> > > Yes, it works perfectly. I was able to use mplayer and fbtv and tvtime
> > > with DirectFB and/or framebuffer. I also can use a X-Server with this
> > > card. 
> > 
> > Good to hear. Does your setup work with a vertical refresh rate of 50Hz?
> > 
> Yes. I use the following fb.modes setting for my monitor resp. my 15:9
> LCD-TV:
> 
> mode "1280x1024 50Hz 32bit"
>     geometry 1280 1024 1280 1024 32
>     timings 10707 248 16 144 38 1 3
> endmode
> 
> mode "1280x768 50Hz 32bit"
>     geometry 1280 768 1280 768 32
>     timings 11415 248 16 144 30 1 3
> endmode
> 
> 
> 
> > > To subsume this discussion, i have to do everything in software (with
> > > non CLE266 hardware) to get a 'high quality' solution. 
> > 
> > I think yes. But even on a low-end CPU (VIA C3 1Ghz, Athlon 800Mhz) you 
> > only need about of the 50% CPU power for pure audio + video decoding.
> > 
> > > I will think  a
> > > little bit more about my last point (using movement vectors). Could be a
> > > good starting point for working with mpeg2.
> > 
> > You would have to break up the decoding functions of libavcodec to get 
> > the required informations, unless you don't want to parse the whole mpeg 
> > stream twice. I don't say it's impossible, but you have a long way to 
> > go. 8-) You could write a standalone tool that simply takes a video pes 
> > stream, decodes and displays it -- without all the audio stuff and a/v 
> > sync code. This should let you concentrate on the important things.
> > 
> > > Do you think it would be
> > > interesting to add deinterlacing support to your softmpeg library, and
> > > is it possible to support your work on the library.
> > 
> > Sure. Currently, "deinterlacing" on the "local" side (ie. not on the tv 
> > out) is done in the sense of "stretching every field to the full frame 
> > size and blitting every 20ms".
> > 
> > I admit that this is currently more or less hardcoded for Matrox and 
> > cle266 and might not work for other adapters. (which aren't support by 
> > DirectFB anyway *sigh*)
> > 
> 
> Does this mean, you stretch one field (288 lines) to a full frame
> without weaving two consecutive fields together for one frame. Wouldn't
> it be easy to do at least some kind of weaving, which is perfect for
> film mode material.
> 

If you stretch all the fields to the full size of frame, and show them at
50/60 fps, you get perfectly smooth video (if the content is interlaced).

Blending the fields together and showing as single frame drops the "fps" to
the half.

tvtime(.sf.net) does "full-framerate" deinterlacing using many algorithms
under Linux..

> > We currently use triple buffering for video display and we don't buffer 
> > decoded frames. It depends on the deinterlacing algorithm if you need to 
> > access "older" frames, so you might need to implement decoded frame 
> > buffering. The current logic does "just in time" decoding, then 
> > memcpy()s the frame to the destination surface, then Flip()s it. This is 
> > efficient and work very well to achieve good av sync.
> > 
> > On the cle266, the tv-out is programmed to output interlaced. I can n 
> > confirm that it can output progressive, too, but we don't have means to 
> > implement and test this currently.
> > 
> > As you can see, "libsoftmpeg" is performing quite well, but is 
> > work-in-progress.
> 
> so i will have a look into it, to start to implement the things i would
> like to see and perhaps if they are good enough they will make their way
> back to the library.
> 
> Thanks
> michael plagge
> 
> 

-- Pasi Kärkkäinen
       
                                   ^
                                .     .
                                 Linux
                              /    -    \
                             Choice.of.the
                           .Next.Generation.


-- 
Info:
To unsubscribe send a mail to ecartis@linuxtv.org with "unsubscribe linux-dvb" as subject.



Home | Main Index | Thread Index