> Richard Rauch said:
 [...]
> > Remember that 4 bits per channel != 4 levels, and 12bpp != 12 colors.
> >
> > Calling 12bpp "severely posterized", I think, is a bit extreme.  (^&
>
> Not at all.  I use my pc as a DVD player to feed a projector which hits a
> screen about 6 ft wide.  At this size, the difference from 30bpp to 24bpp is
> apparant.  When viewing on a 19" monitor which has been calibrated, the
> difference from 24bpp to 16bpp is very apparant.  Trust me, if you see a well
> setup DVD player on a PC, many things become apparant - the exactness of the
> PC unmasks many faults of the system.
> Again, I'm picky.  But these are realworld "issues" - it's not a blue stripe
> around the outside of a CD...
> If 12bpp works for you, then go for it.  4 bits/channel = 16 distinct
> shades/channel (of course, this assumes an even color distribution, which is
> reasonable but not always realworld).  Trust me, seeing 16 shades of green
> trying to represent a big, green field just doesn't work.  It looks like a
> 1970's psychedelic poster.

A field composed entirely of shades of pure green sounds like modern art.
As far as I'm concerned, you can't do too much damage to that.  (^&  (The
point, of course, is that real life "green" is not just a linear gradation
from black to pure, fully saturated, full-value "green".)

Anyway, I've not seen your 6' screen so I can't comment on it I guess.
However, on my setup (at least with Blade Runner as my sample), ``mplayer
-framedrop -double -dvd-device /dev/rcd1a -dvd 1'' has only two noticable
shortcomings:

 * It has to drop frames to keep up.  (A bandwidth issue.)

 * The resoultion is fairly dinky (720x480).  (Of course, higher
   resolution would aggravate the bandwidth---OTOH, lower resolution might
   give a viewable movie without frame dropping.  But, I'd really like to
   stretch it out to my normal X server's full width.)

It may be that the film, the processing, or what was actually in front of
the camera provided a natural graininess that effectively makes for good
dithering.  I may find that some other films (especially animated ones or
colorized ones (if I should stoop so low; (^&))  have overly flat colors
that cause the banding/posterizing effect to become visible.


(I should also say one other thing...I was *sure* that it said it was
using 12bpp before.  When I was testing Frederick's suggestion for
postprocessing, all I could see were acknowledgements that it was on a
16bpp display.  Maybe it was ogle cutting back to 12bpp?  Maybe it was
when I was on a 24bpp X server?  Maybe I just dreamed it all...(^&)


> > I suspect you've not actually used it, or do not appreciate how
> > significant the variation between mplayer and ogle is---or overestimate the
> > difference between 12bpp and 16bpp (or even 12bpp and 24bpp).  If
> > represented in 12bpp, Ogle's "black" would be about 222 on my system, I
> > think.  mplayer's black is black (000).
>
> Again, I've used ogle [albeit briefly, say for 45 minutes] on my 6 ft wide
> front projection screen.  It's definitely got room for improvement, but it's
> not _bad_.  I have not used mplayer - I can't comment there.

Well, not terrible, no.  But maybe a little worse than mplayer's output.
All things are relative.


> > If the server is running in 24bpp, there are artifacts (most visible when
> > the image changes rapidly, as when a beam of light moves rapidly over an
> > object).  This causes visible tearing.  At 16bpp, the images don't degrade
> > (since they are 12bpp anyway) and the tearing goes away, being replaced by
> > smoother animation.
>
> That's a possible effect of posterization (not enough color space to
> accurately display detail).  It could also be inadiquate bandwidth to the card

I considered both possibilities, but if it's double-buffered you shouldn't
see any such tearing.  (The idea, as I understand it, of a double buffer
is you display one image while you render the other.  Then during the
vertical blank of the video refresh, you swap the two.  You should never
see tearing if you double-buffer.  Unless they totally screwed up
double-buffering, that is...(^&)

Bandwidth was my first thought, though, and it explains why a single
buffered display wouldn't be able to handle it without tearing.


> (24bpp == more data == more dropped frames == artifacts).  If you want to
> avoid the tearing and combing, then you have to deinterlace the image.  Most

What is "combing"?  I've never heard this term before and hadn't noticed
anything in playback of which the term is suggestive.

Or is combing just another term for frame dropping?  I guess that that's
vaguely suggestive...  The frame dropping is clearly a bandwidth issue,
and is livable for now.  (It will be one more cause for joy when I get a
DRI-supported AGP card "one of these days", though.)


> > I'm not sure what constitutes "getting serious".  A movie *is*
> > entertainment, after all.  (^&  (And I don't plan on charging admission to
> > let people watch movies on my computer.)  However, I'm satisfied enough
> > that I'll probably buy a movie from time to time.
>
> Well, that's fine for you.  Many people will argue that music is entertainment
> as well, yet there are people out there willing to spend tens of thousands of
> dollars on good loudspeakers, amps, etc...  Same thing for wines, cars and
> computers...

I was just poking at the idea of refering to a pursuit of entertainment as
"serious".  (^&  And, too, suggesting that there's some relative judgement
in there.

If you're a consultant pulling down $100K a year, then a really nice
DVD playback system may be a cheap toy for you.  If you're a grad student
on a $12K/year stipend, then a "serious" interest probably has a much
lower ceiling on what you'll put into it.  (^&

(But, no, even relative to my level of income, I'm not serious about it.
It's cool that I can watch movies on my computer, and I care about their
visual quality.  But I'm content to stop spending money on movie playback
support for now.)


> >> difference.  But, then again, I'm picky about my video.
> >
> > Heh.  Most are, I think.  The question is the relative weight that <X>
> > enjoys versus <Y>.  I'm a grad student with very low income at the moment,
> > so a $400 or $500 video card just on which to watch a handful of movies is
> > a huge waste.  A $40 or $60 card that I'll "have" to replace in time with a
> > "real" card for DRI support is also a waste.  A $100 or $150 card when DRI
> > is available is almost defensible, though, so I'm waiting...the longer I
> > wait, the more leverage I get for that $100 or so.
>
> Absolutely.  But realize that if you're going to be bothered by some of what
> you've seen, it will either be solved by 1) hacking some good filter code into
> ogle/mplayer and/or 2) upgrading to an AGP card.

Well, I think I've said this, and you've said this, and others have said
this.  (^&

And, I'm not really bothered by anything in these programs at this point,
except Ogle's poor color handling, mplayer's man-page (things are not
sorted very well) and both programs' frame dropping.  And the dropping is
just a bandwidth issue which I fully expect will "magically go away" when
I eventually buy a real video card, sometime down the road.  (^&

Things are okay, now.  Not great, but okay.  If I had all of Bill Gates's
money, I might be better off.  But with my financial resources being what
they are...  It will do.


> P.S. if you're ever around NYC and want to see some of what I've been
> soapboxing about, do get in touch.  We'll grab a beer and see some movies.
> Maybe we can get some of the other netbsd-nyc people over for a good film.

It's a little out of my way (I'm in Houston, TX), though I may take the
time to go to some conferences this year.  (One perq about being a grad
student here is they put aside a small budget to send us to
conferences...(^&)


  ``I probably don't know what I'm talking about.'' --rauch@math.rice.edu