[mythtv] Filter interface change proposal, working on patch.

D Banerjee davatar at comcast.net
Thu Oct 23 10:59:12 EDT 2003


----- Original Message -----
From: "Edward Wildgoose" <Edward.Wildgoose at FRMHedge.com>
To: "Development of mythtv" <mythtv-dev at mythtv.org>
Sent: Thursday, October 23, 2003 6:09 AM
Subject: RE: [mythtv] Filter interface change proposal, working on patch.

> However, for sure, the effort required to decode is significant with HDTV,
it's a VERY big picture

Yes, that's the point, unless you're going to get two or more bleeding edge
machines. I plan to have a bttv card and a pchdtv card in one machine, which
will also be a frontend.

> I would tentatively suggest that it would be worth optimising filters
primarily towards the mpeg cards.  The point is that those who are looking
for high quality output are likely to be using one of these already (ie DVB,
HDTV, and PVR250 users).  People with bttv cards will already be using a
pretty high spec machine and are likely to be running out of juice on all
reasonable hardware (as of today) if trying to do 720x576 mpeg4 recordings,
plus filters on the playback.  Just a thought.

This is inherently a bad idea. Filters that aren't applied to a raw stream
before lossy compression are really just a kludge. In fact as far as
filtering noise goes, if you don't do it before encode you end up with
macro-block noise, not the original noise, which means you've already lost
alot of 'real' image. Then you need to apply filters to get rid of that,
etc, etc. HDTV and DVB don't need filtering, they are raw mpeg streams, no
additional noise. As far as playback filters go, if you've done your
encoding job correctly, you shouldn't need it. On the other hand if you've
encoded at some rate like 300MB/hour (which works remarkably well for the
simpsons and sttng) you might want something to clear up the macroblocks.

The highest quality analog encoders will be done by software, as the raw
stream can be processed correctly, and then encoded at a high bitrate with
good compression like mpeg4. Mpeg2 just isn't that good unless you're going
to be using 4,8,12GB an hour (On noisy broadcast video). Considering I'm
recording 480x480 mpeg at 3000mbps with quickdnr on a crusty old athlon 800,
I don't see this as a problem. (the noise reductions _help_ cpu). Btw I was
dead set on getting a pvr-350, now not so much because I've seen the kind of
quality I can get with bttv, with reasonable disk usage, direct encoding. To
get the same quality with a hardware encoder (if possible) I'd have the
crank the bitrate up as high as it goes on record and then postprocess,
which starts to lose it's attractiveness when you start recording many shows
in one day, each one taking 2-3 hours to transcode (If not more because hdtv
is being decoded, and filters) It would be easy to fill the disk. Not to
mention lossy->lossy gives you a IQ hit right there.

I think the main users of pvr-250's will be people who want many tuners in a
backend or lightweight backends. Once the filtering interface is running
properly anyway. Besides there are yet-to-be-implemented filters using 3d
hardware acceleration, like ati all-in-wonders do (or claim to) on windows.

Debabrata





More information about the mythtv-dev mailing list