[acannon]

Now that the TheaterTek DVD player has been announced (yeah!) and updates to PowerDVD etc. are starting to roll out, I was wondering what the prospects are for improved deinterlacing in future software players? This has issue has been skirted around in a number of threads (especially with the recent SDI/DScaler alternative being introduced).


The features that are now being provided by software players are very neat and are compelling reasons to buy in themselves (I'm on the pre-order list for TT). Still, there seems to be a gap in terms of deinterlacing quality. The discussions on mating DScaler with output from software DVD players makes for interesting reading, but I guess I don't know enough to fully understand why a developer hasn't yet gone that extra mile. Is it a fundamental problem with the way DVD playback is handled by the OS, or is a question of not enough processing power, or...?


---

 

[RyanDinan]

I've been wondering the same thing, and even asked this question in the TheaterTek DVD player thread.

The lack of 3-2 pulldown detection as a de-interlacing option is one very big feature that has been missing from software DVD players.

I would have thought that by now, it would be a standard feature on most. Apparently not.

Since there seem to be quite a few DVD's out there that are poorly flagged, they would look so much better with a cadence detection scheme, instead of a forced bob/weave (which is the only option available in current DVD players). The one exception to this, is ATI's player, which will use adaptive de-interlacing, which is a kind of "combo" bob/weave, and yields much better results.

I just don't understand why it hasn't been implemented yet. Obviously, it's a more intensive de-interlacing scheme, but nowhere near as intensive as adaptive de-interlacing.

There's got to be enough cpu horsepower left over to do this! My AMD Duron never uses more than 30% when playing back a DVD. That other 70% should be more than anaquate to handle 3-2 pulldown detection. Essentially, all that would have to be done is read a few fields ahead, lock onto the 3-2 cadence (if present) and weave the correct fields together. I'm sure it's more involved that that, but I just don't see it being beyond current cpu's abilities.


What do you think?


-Ryan Dinan

 

[acannon]

It sounds like we're thinking along the same lines. It doesn't seem like a better deinterlacing scheme than forced bob or weave would be too, too difficult (or costly, in terms of processing), especially given the precedent set by the ATI player.


I'd love to have the flexibility and power of the deinterlacing algorithms used by DScaler integrated into a player, but realize that may not be realistic... even something quite simple would be an improvement.


---