[opendtv] Re: Natural Motion on your PC
- From: jeroen.stessen@xxxxxxxxxxx
- To: opendtv@xxxxxxxxxxxxx
- Date: Thu, 10 Jun 2004 15:22:06 +0200
Hello,
I wrote:
>> It depends on what you mean with "duplicate" ?
Doug McDonald <dtvmcdonald at yahoo dot com> wrote:
> I mean, IN PRINCIPLE, to design a TV system to do it.
(It depends on what you mean with "a TV system"...)
> What I want to know is what it does! You have never
> told us that.
Well, you could have looked up one of those countless
publications by Gerard de Haan, including his PhD thesis
and the multimedia book that he wrote and distributed.
That contains more secrets than I could ever explain.
But basically, what it does is:
- motion estimation based on recursive block matching,
- motion interpolation (reduce the motion vectors to
the new temporal positions between original frames),
- up-conversion (move the video along those vectors to
the new spatial position according to the desired
temporal position).
Once the repeated temporal positions from film material
with 2:2 or 3:2 pull-down is properly recognised, then
removal of the film judder is rather trivial.
Recursive de-interlacing is also based on motion
compensation: the new frame is a mix of a new field
and a motion-compensated previous frame.
> Does it, for example, require a special 72 Hz
> progressive TV set so it can just replay every
> frame (deinterlaced of course) three times?
The software implementation under discussion requires a
computer monitor, not a TV. 72 Hz would be a good choice.
And no, it does not do just frame repetition (i.e. 3:3
pull-down). That would be way too easy.
> Does it actually do actual motion interpolation
Yes !
> (Which would need the processing power of perhaps 50
> 2.4 gHz Pentiums to do properly in real time).
That is the beauty of it: because this is only for SD
resolution, and because of heavy code optimisation, this
algorithm runs easily on a single (2 GHz ?) Pentium 4.
Today I have seen it run live on a modest PC in this lab.
The only complaint I've heard was that it required the
latest WM9b player from MS, which makes some PCs unstable.
You can download a trial version of this WinDVD for free.
and there's even rumour of a crack being available
> The problem with this is that "24 Hz judder" is
> something that appears not to be a problem at all.
Me thinks that a minority of people would think that...
> I saw a Philips high-end set in a store that
> apparently had it. I say apparently because it didn't
> say much.
A good demo speaks volumes. Maybe it was a bad demo ?
Of course every technology comes with its own artefacts,
but on the whole this is still a vast improvement.
> This set certainly was no great winner ... certainly
> whatever it did was not as good as the Faroudja chip
> in my own TV.
Faroudja does a good motion-adaptive (NOT motion-
compensated) de-interlacing, but not much else.
No motion-compensated frame-rate (up-)conversion.
> It is not clear what this set does with DVD
> 24 Hz film material. It sure looks great, however, but
> nowhere near as great as 24 Hz film material
> upconverted to 720@60p by our local Fox station, which
> is simply stellar.
At a stellar equipment price, no doubt.
If every station transmitted clean 60p, our jobs here
would be mostly over... :-(
> I think you and anybody who worries about something
> called "24 Hz judder" is barking up an empty tree ....
> it's a non-problem.
Once you have seen the difference, you'll know what a big
problem it is. In fact, you have just implied it yourself,
by stating that the upconversion to 60p by Fox is stellar.
So the original 24p was less than stellar ?! QED.
> Europeans have chosen a system to try to cure a
> problem that would cure itself with time (i.e.
> multipath) and in the process lost a lot
I will believe you when the number of happy customers in
the US has overtaken the number of ditto in Europe...
Don't forget that multi-carrier was invented by your MIT.
It prevents a problem from occurring in the first place.
But I'm not supposed to talk about this. So I don't.
Greetings,
-- Jeroen.
|-----------------------------+---------------------------------------|
| From: Jeroen H. Stessen | E-mail: Jeroen.Stessen@xxxxxxxxxxx |
|-----------------------------+---------------------------------------|
| Building: SFJ-5.22 Eindhoven| Philips Digital Systems Laboratories |
|-----------------------------+---------------------------------------|
| Phone: ++31.40.27.32739 | Visiting & mail address: Glaslaan 2 |
|-----------------------------+---------------------------------------|
| Fax: ++31.40.27.32572 | NL 5616 LW Eindhoven, the Netherlands |
|-----------------------------+---------------------------------------|
| Pager: ++31.6.6513.3818 | Visit us: http://www.pdsl.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:
