View Full Version : Some clarification please
aris
16th September 2002, 08:26
I think I have this figured out, but would like confirmation..
Since I view my movies on a PC with a TV-OUT, from a 640x480 Resolution, I've noticed that 640x480 movies will both look better qualitywise and tax the CPU significantly less than other-resolution movies, as they don't require real-time resize by the player app (WMP in my case). Effectively a 640xN resolution movie will require less CPU and look better then the same DivX resized to 650x or 630x during encoding.
That said, I do all movies at 640xN (N goes according to the DVD's AR).
The logical yet annoying part is that CUT(Pan'N'Scanned) movies that come at 4:3 AR on the DVD, which, for the reason stated above, I'll want to resize to 640xN=480 resolution during encoding, will effectively take up twice as much space after being resized (blown up), as they have practically twice as many pixels covering the screen than a 1:2.35 would have. And to achieve my desired quality - will take 3x700 CD's. That's >2Gigs per 4:3 movie on my fileserver. Yuck.
Can anyone shed some more insight, prove my 640x guideline as futile or confirm its relevance, or in any other ways give useful advice on this issue?
TheWEF
16th September 2002, 21:36
Originally posted by aris
I've noticed that 640x480 movies will both look better qualitywise and tax the CPU significantly less than other-resolution movies, as they don't require real-time resize by the player app (WMP in my case).
something is wrong with your gfx-card setup. resizing should be processed in hardware and not effect cpu-load.
if your hardware overlay is not working correctly you usually get a very bad picture quality. try different desktop color-dephth or new gfx-driver.
wef.
aris
17th September 2002, 08:44
What resolutions are recommended?
(Rather, what's the highest resolution I can use that will still provide quality improvement? Assuming that if I give the DivX a higher res than the origional, It won't improve anything, just eat up more space)
And, given you answer the last Q with a number,is that the resolution of a half-frame in an interlaced stream, or of both halves put together?
I assume that what you say is true, as you obviously have more experience in the field than I do, and yet the picture still looks better on a 640x DivX than on an any-other-res that's been resized by my GFX card. Apparently, hardware or sw-inside-encoder, there's still different resize algos and different outcome quality...
TheWEF
17th September 2002, 22:44
Originally posted by aris
What resolutions are recommended?
desktop res? - well, what looks best for you is best for you.
watching a movie of 640x... on a 640x480 desktop can't be wrong i guess.
i would not recommend 1280x1024 because pixels are not square - you get wrong ar.
i use 1600x1200, 32bit on my 22inch and i love it.... :)
wef.
aris
18th September 2002, 10:15
What I meant was:
What resolutions do you recommend using to encode the DivX5 video stream? :-)
For 1:1.66 through 1:2.35 movies (with everything inbetween) I'm pretty happy with the 640x tacic, both in space demand and qualitywise. Then watch it on 640x desktop, and it looks terrific.
I'm asking because I still haven't figured out some optimal tactic for the Pan&Scanned 4:3 versions.
AND, while at it, another Q:
I have an interlaced 4:3 25fps PAL movie, and when I do a standard by-doom9's-guide DivX5 2pass avi,
I keep getting a result where many objects are "half-transparent", *not* as if it wasn't deinterlaced, it was (I actually tried with both deinterlace methods, both with the same result) but as if you're looking at an object with both your eyes, then cover one eye with your hand and you see your hand semi-transparent. That kind of semi-transparent. For example, someone is waving his hand in the movie, and when I pause a specific frame, the hand is semi-transparent. No horisontal-lines smudge you'd get on a broken deinterlace. Then again, maybe that's just what it is.
Another possibly-useful piece of info is the fact that it's a 30-year-old classic movie made in Israel (not exactly hollywood), And the DVD I bought says it's compiled from the original master. I tried getting a similar result from the VOB files but they don't seem to have this semi-transparent issue.
Another possibly-useful fact is that in preview, the picture looks like crap. It's blocky (relevant?) and the frames are interlaced.
Further, I made sure "force film" in Field-Operation in DVD2AVI is *OFF* when I created the dvd2avi project (I understand it's only used for NTSC FILM sources).
Any ideas what could be causing this and how to overcome it?
Thanks!
manono
19th September 2002, 00:29
Hi-
If you're happy with your 640*xxx movies as they are, then keep using it. Optimal is what looks good to your eyes. You're doing this for yourself after all (and maybe to impress your friends). Personally though, I'll run a compression test first to help choose the resolution.
I'm asking because I still haven't figured out some optimal tactic for the Pan&Scanned 4:3 versions.
4:3 movies are much harder to compress for the same horizontal resolution (in your case, 640) because when compared to 2.35:1 movies, there are nearly twice as many pixels. So, to keep the same quality, you'd have to lower the resolution, or make it for an extra CD. But the optimal tactic is not to get them at all. That is, when given the choice between the widescreen version and the pan and scam...er...scan version, always choose the widescreen version. Go check out Widescreen.org (http://www.widescreen.org/widescreen.shtml) and then decide if you still prefer the butchered version. But, of course sometimes you don't have a choice, and many movies were actually created at 1.33:1.
As for the half-transparent figures in your movie, that's what the blending style of deinterlacing does-they're sometimes called ghost images. Since many PAL movies just have the fields shifted, you might try and use Telecide() by itself. If it works, you won't need to deinterlace at all. If it doesn't work, you might try FieldDeinterlace(Blend=False) to see if you prefer interpolation instead. Refer to the Decomb help file included in GKnot for more details.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.