Lanczos vs Lanczos4

[lpvtech]

The title says it all... Which is better? For what kind of sources are they better? What should I use for a normal encode... that for example to encode my "The Lord Of The Rings" DVD to about a 2700MB XviD Rip.. I prefer a resolution of about 720x304..
Since I don't want to lose any details..

Thanks in advance...

[graysky]

Well, first off, good for you for selecting a mod16 resolution.

Secondly, unless you have a hardware divx player, why are you encoding to xvid? x264 gives superior results in my experience.

Finally, my understanding of Lanczos and L4 is that both are a "sharp" means to resize and that having artificially sharpened images make the encoding tough without artifacts at low bitrates. What is a "low" bitrate? I dunno, I read about this a while ago and can't find the urls to give you. I found that using the splines36 resize was a better option since it more neutral to the resulting pixels.

Ah, a quick google search gave this (http://forum.doom9.org/showthread.php?t=112594) thread. Let us know what you find out.

[lpvtech]

Thanks.. But I did do a little bit of tests on x264...
But for LOTR XviD definitely seems better.. x264 doesn't do those ultra-fast motion frames prettly well...XviD is the way to go... Also x264 awfully takes too much of time..I checked a sample clip with high motion about 8mins.

I got a 100mb at 1500kbps bitrate (2 pass)
Both XviD and x264 are of exactly the same quality..I even checked pixel by pixels...Spend a lot of hours comparing.. They were totally the same... except one thing... While using x264 the seeking is slow.. and I get a feeling of flicker or like very minute pauses if a few milliseconds often.. But XviD is just perfect..
Also XviD took about 8 mins for 1-pass and about 25 mins for second..
But x264 took 10 mins for the 1st and about 1 hr 15 mins for the second...


Options:

XviD : Q Pix, No Adaptive, No GMC, No PAcked,No Interlaced, Keyframes 240, All other left to default with MPEG Matrix.
x264.. Turned on all options except DCT Decimate.. B-Frames: 3
Ref Frames :3 Mixed Ref. RDO(Partial decision: 6)

[graysky]

Well, use whatever you like. I did my comparisons using the HQ-slow profile in MeGUI with one change: the inloop filter settings 0,0 from the stock -2,-1. When I compared xvid @ 800 kbps to x264 @ 800 kbps the differences were clear.

What did you find on the Lanczos vs. splines36?

[davidhorman]

The title says it all... Which is better?

There's only one way to settle it... fiiiiiiiiiiiight!

Somebody, somewhere, will get this joke.

David

[Terranigma]

lanczos is better due to the fact that it's more flexible.
for lanczos4, you'd just add taps=4 to the end of lanczos
e.g.
lanczosresize(720,480,taps=4) #Lanczos4Resize

for sinc256, you'd add taps=8.

But I personally don't use lanczos. I use the spline resizers instead.

[lpvtech]

Thanks... But I don't see any difference between Lanczos and Spline... They both look precisely the same...

[foxyshadis]

Spotting differences when downsizing can take a trained eye sometimes, especially if it's not very much. The differences are mainly obvious with large ratio or sizing up. The difference will always show up in the final encode size though.

[Leak]

Thanks... But I don't see any difference between Lanczos and Spline... They both look precisely the same...
Try interleaving both versions, then step back and forth between both versions of a frame in VirtualDub. That's probably the easiest way to spot such small differences - if they can be spotted at all... :)

np: Boards of Canada - June 9th (Hi Scores)

[lpvtech]

I tried them... It is clear while upsizing....Thanks... But I'll settle with Lanczos while downsizing since it is a little faster than spline due to its simpler algorithm..

[Terranigma]

But I'll settle with Lanczos while downsizing since it is a little faster than spline due to its simpler algorithm..

What's the ratio, in terms of a speed difference, is there between using a splineresizer and lanczos? There's not that much of a difference for me. :)

[lpvtech]

I'm not sure... But its like a about 1.5 to 2 frames faster per second.. I tried it a several times.. The difference is very clear while encoding at a very heavy load.. I mulitask just too much.... All with just a centrino 1.6 single core....