View Full Version : Encoding resolution choice
Chromanin
24th July 2006, 20:47
Heya, I've ripped DVD and am planning on converting it to lossy x264 compressed.
Now I'm simply wondering if there would be any advantage at upsampling the resolution to HD. I've read through the forum and noticed how some people say there's a huge difference, and then I read another reply saying that upsampling is going to lose more detail than what it's worth.
I'm a little confused in this regard, but to be more specific, this is concerning Anime interlaced footage.
*.mp4 guy
24th July 2006, 21:43
You can get some really nice results on anime with slower-then-realtime resizers, it all comes down to what looks better to you, how long you are wiling to wait, and how much space you are willing to use. What is your destination medium? Is time an issue? How detailed is the anime/how is the video quality already (a raw clip would be the most useful answer to this question)?
Lokean
24th July 2006, 21:50
Unless for some reason your player absolutely requires it to be a certain size, I don't see how there's any advantage at upsizing during encoding. But... I'm always willing to learn... ;)
Awatef
24th July 2006, 22:20
I would definitely disapprove an upsizing.
You see, if I have a 320x240 clip and want to encode it to DVD, I don't go all the way up to 720x576, I use the smallest possible size, being 352x288, because: (1) I'm not creating details; (2) I'll be wasting bitrate.
In your case, you don't even have standard constraints, so if you're compressing to H.264 and want maximum detail, keep the original DVD resolution, and push your bitrate to the highest level you can afford, and be happy if you don't get too many artefacts ;)
Chromanin
24th July 2006, 22:43
hmmmm
The thing is tho, that watching a very low res video file on a very high res screen is going to severely blur things out. So I'm guessing that upsampling the same image and giving it a strong sharpen would make the IQ much better, am I right? The big screen thing concerns me, because I'm planning on using it for a pretty big screen, even if DVD quality is much better that 320x240.
*.mp4 guy
24th July 2006, 22:49
mpeg2source("Steam Boy.d2v").crop(2,6,716,464)
Lanczosresize(1280, 720) (http://img374.imageshack.us/img374/6938/lanczostd1.png)
mpeg2source("Steam Boy.d2v").crop(2,6,716,464)
Degrainmedian(mode=3)
MultiSWAR(1440, 784, smoothing=true, Sharpen=16, NthPass=true, Sthresh=10, Sstrength=1, StepRatio=4, warping=1)
Spline36resize(2880, 1536).Xsharpen(110, 255).Xsharpen(110,255).blur(1.42).blur(1.42)
Spline16Resize(1280, 720)
Degrainmedian(Mode=3) (http://img470.imageshack.us/img470/9241/multiswarpz6.png)
foxyshadis
25th July 2006, 03:25
"Spline36resize(2880, 1536).Xsharpen(110, 255).Xsharpen(110,255).blur(1.42).blur(1.42)"
Now that's a little silly, you'd get a similar effect with something like
LimitedSharpenFaster(smode=4,strength=200,special=true,wide=true,soft=20)
in a quarter of the time.
If you can afford both the bitrate and the time and you really want your incredibly sharp video in realtime, upsizing is a no-brainer.
*.mp4 guy
25th July 2006, 03:58
Its actually faster then Limitedsharpen plus 1.5X supersampling (Limitedsharpen Looks very ugly to me without supersampling), I never looked into limitedsharpenfaster, how is the quality/speed trade off?
Chromanin
25th July 2006, 04:17
Thanks alot peeps, I'll try out/experiment with all of the methods suggested and see what suits me best :).
foxyshadis
25th July 2006, 04:22
Smode 4 uses only 1.25x SS, by default. And it's at around twice as fast as the old LS, if not more (though I haven't measured in ages).
sysKin
25th July 2006, 04:41
I would definitely disapprove an upsizing.
You see, if I have a 320x240 clip and want to encode it to DVD, I don't go all the way up to 720x576, I use the smallest possible size, being 352x288, because: (1) I'm not creating details; (2) I'll be wasting bitrate.
That is 100% true for mpeg2 or asp encoding. The smooth surfaces created by upsizing have no new details (obviously) but use a lot of bitrate, because smooth gradients still have many DCT coefficients.
However, this might not be true for avc encoding. It has inloop deblocker and complex intra prediction, both of which handle smooth gradients very well with little bitrate.
IMHO it's worth experimenting with x264.
Chainmax
25th July 2006, 04:53
Smode 4 uses only 1.25x SS, by default.
...
Really? Then why is it recommended to use Soothe when using SMode=4?
foxyshadis
25th July 2006, 05:54
Supersampling doesn't have anything to do with temporal stability, it's for anti-aliasing or more precise sharpening mainly. Seesaw, for instance, uses no supersampling yet has to integrate Soothe or the interframe jitter can be very distracting. (You can try it for yourself.)
Didée
25th July 2006, 09:43
@ *.mp4 guy
Without trying out, I'm absolutely positive that LimitedSharpenFaster runs faster, and not only marginally ...
BTW, there's no speed/quality tradeoff between LS and LSF. It's only about speed, quality is exactly the same.
@ all upsampling-disbelievers
Wasn't the story about upsampling how/when/why and hownot/whennot/whynot told often enough already? I'm too lazy to repeat myself over & over again ... search for "upsampling" by "Didée" display as "posts" ...
For ex., a higher resolution source vs. an upsized lower-resolution source (http://forum.doom9.org/showthread.php?p=841552#post841552).
(Note that this is a raw 4500 kbps DVB-Source vs. an already encoded 1400 kbps XviD.)
What do you think would have come out if just the ananamorphic source, without any filtering "that would only make things worse", would have been encoded? :)
check
25th July 2006, 11:19
re: searching for your thread, damn your accented character didée!
Didée
25th July 2006, 11:51
No problem ... just type "Did", then a selection of users starting with those letters should pop up ... or for that matter, just leave it away ... will work out, too.
Immediately, the thread "Anything to gain by using higher resolution than the DVD (http://forum.doom9.org/showthread.php?t=85220&)" springs to my mind.
vBulletin® v3.8.5, Copyright ©2000-2012, Jelsoft Enterprises Ltd.