[opendtv] Re: De-interlacing with HQV high quality video processing
- From: Jeroen Stessen <jeroen.stessen@xxxxxxxxxxx>
- To: opendtv@xxxxxxxxxxxxx
- Date: Thu, 6 Oct 2005 13:12:10 +0200
Hello,
It's interesting that some of you guys have seen errors in
the output from that de-interlacer. I haven't personally seen
it (yet). Earlier this year I have had a demo of National
Semiconductors' latest design, and that was quite impressive.
Significantly better than Faroudja's DCDi, which is still the
quality standard to compare with on today's market.
Of course any de-interlacer can be made to fail with the
"proper" signal: undersampled, no diagonals to take hints from,
and vertical object motion at the critical speed.
But I will counter that with three arguments:
- interlacing _should_ be used as an oversampling technique,
not as an undersampling technique,
- the really critical cases occur only rarely, and
- an all-progressive system can show the same interlace
artefacts at its own critical (vertical motion) speed.
So similar restrictions apply to progressive systems, they
should also not be undersampled !
Dan Grimes wrote:
> While I find interlacing reprehensible, since there are interlaced
formats
> that need to be displayed, couldn't displays be designed to display
> appropriate pixels at the appropriate time?
Yes and no. For a light valve display (LCD etc.), not lighting
up half of the lines means wasting half of the lamp power.
Only when you physically move the image, as in the Wobulation
DLP display by HP, can you make an efficient interlaced light-
valve display.
For emissive displays it is relatively easier to just blank
half of the lines and thus obtain an interlaced raster. It does
not necessarily degrade the efficiency, but it will degrade the
peak light output power by 50% (i.e. less "sparkle").
It gets better if you can modify the actual scanning raster.
In the CRT this is traditionally done by modulating the vertical
deflection current with a small offset (time or position). For
a laser beam display (incl. GLV) one could do the same thing.
Finally, the ALiS plasma display from Fujitsu-Hitachi is a true
interlaced plasma display (1024i), but it can also be applied
as a progressive display (512p). The latter is inferior...
A later design manages to achieve full progressive resolution
from the same reduced amount of row drivers, and then it would
seem unwise to degrade it to interlaced again (50% peak output).
> Seems to me that it is just a matter of clocking in the right
> pixels at the right time, allowing a progressive or interlaced
> image to be displayed natively.
No... you have to think about whether you are actually using
all of your resources efficiently for making light output.
The addressing is then merely a minor technical challenge.
Simply generating 50% of the data as black is good enough.
> Of course, it doesn't account for the different resolutions.
Right, and there is the bigger catch: (vertical) scaling can
ONLY be done on de-interlaced signals. So in practice de-
interlacing is always necessary. And then you might as well
send it to a progressive display. Unless you have a display
that performs better in interlaced mode (CRT, ALiS), then at
the end you would convert the signal back to interlaced again.
I still believe in interlacing, as a form of an offset
(quincunx) sampling grid, because it puts the highest
resolution where or when it is needed the most. This holds
for cameras, displays, and transmitted data. And there are
now several examples of good-enough de-interlacing circuits,
so that one can no longer say that the problem is unsolvable.
And of course my company would not permit me to say otherwise,
even IF I would think otherwise. Which I don't.
Greetings,
-- Jeroen
+-------------------------------+------------------------------------------+
| From: Jeroen H. Stessen | E-mail: Jeroen.Stessen@xxxxxxxxxxx |
| Building: SFJ-5.22 Eindhoven | Deptmt.: Philips Applied Technologies |
| Phone: ++31.40.2732739 | Visiting & mail address: Glaslaan 2 |
| Mobile: ++31.6.44680021 | NL 5616 LW Eindhoven, the Netherlands |
| Pager: ++31.6.65133818 | Website: http://www.apptech.philips.com/
|
+-------------------------------+------------------------------------------+
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:
- Using the UNSUBSCRIBE command in your user configuration settings at
FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
unsubscribe in the subject line.
Other related posts: