2.07 or 2.5 which way would u go for dvd to dvd

[benf2]

I know its late in the game for asking this question. But i still get confused on the yv vs yuy issue. I do all my movies with no plugins...just the basics load, source, kill audio then import to cce with burning to dvd a final result. I guess what i am asking is 2.5 or 2.07 the better(faster & color quality) for my needs.... Played with both seems 2.07 ran faster for me(speeds around 1.7 compared to 1.4 with 2.5).

I went thru 11 pages of search but the yv and yuy ...eeek

what do u recommend when dvd is the source and output?

[primitive]

Originally posted by benf2
I know its late in the game for asking this question. But i still get confused on the yv vs yuy issue. I do all my movies with no plugins...just the basics load, source, kill audio then import to cce with burning to dvd a final result. I guess what i am asking is 2.5 or 2.07 the better(faster & color quality) for my needs.... Played with both seems 2.07 ran faster for me(speeds around 1.7 compared to 1.4 with 2.5).

I went thru 11 pages of search but the yv and yuy ...eeek

what do u recommend when dvd is the source and output?

From what I understand, this is the issue.

Color space conversions are bad for three main reasons: 1) they're slow, 2) they make you lose color information, and 3) filtering operations are slower in some color spaces than they are in others.

DVD video is in YU12, while IIRC CCE accepts RGB32 and YUY2 as input. This means that eventually you're going to have to convert the video stream to YUY2 before passing it to CCE.

With Avisynth 2.5, the plugins you'll primarily use will operate in YU12. This means that any filtering plugins you use before passing the video to CCE will be faster than they would be if you were using Avisynth 2.07.

Avisynth 2.5:
extract video (YU12) --> filter video (YU12) --> ConvertToYUY2() --> CCE

Avisynth 2.07:
extract video (YU12 --> YUY2) --> filter video (YUY2) --> CCE

If you are using filters, Avisynth 2.5 will be faster and more accurate.

(primitive)

P.S. If you're just doing DVD transcoding, you might want to use Avisynth 2.5 anyway: Mpeg2Dec3 has a postprocessing option that you might find attractive, and it's bundled with Avisynth 2.5.

[benf2]

cool.,thanks...it just got a little confusing reading the threads when a lot of people are talking about using it for divx and i am doing dvd

[scmccarthy]

If you had a DVD encoder that could read YV12, 2.5 would be better. I personally don't think it matters which color space the filters work in, so I think it is a toss-up. Using both would give you the flexibility of using either set of filters. The filters could be the deciding factor.

Stephen

[Suzahara]

Originally posted by scmccarthy
If you had a DVD encoder that could read YV12, 2.5 would be better. I personally don't think it matters which color space the filters work in, so I think it is a toss-up. Using both would give you the flexibility of using either set of filters. The filters could be the deciding factor.

Stephen

Well, YV12 filters are actually faster since YV12 is a faster colorspace. Of course if the filter isn't near optimized, doesn't matter then ;)

[scmccarthy]

YV12 is a faster colorspace As I understand it, the speed increase of using YV12 is in avoiding color space conversions. The conversion slows down processing.

Stephen

[primitive]

Originally posted by scmccarthy
As I understand it, the speed increase of using YV12 is in avoiding color space conversions. The conversion slows down processing.

Stephen

Yikes, this is really straining my knowledge.

YU12 is 4:2:0 Y Cr Cb
YUY2 is 4:2:2 Y Cr Cb

YU12 actually contains less pixel data and is therefore fundamentally faster to work with.

[scmccarthy]

YU12 actually contains less pixel data and is therefore fundamentally faster to work with. YV12.

That sounds like an idea that sounds right, but might not be. In light of some people complaining that planes are harder to deal with and others saying it is easier, I imagine it also depends on how the programmer deals with the new color space.

Stephen

[hakko504]

Originally posted by scmccarthy
That sounds like an idea that sounds right, but might not be. In light of some people complaining that planes are harder to deal with and others saying it is easier, I imagine it also depends on how the programmer deals with the new color space.

Stephen I'd say it depends on what you are trying to do. Deinterlacing seems to be hopeless in YV12 as you must jump between planes for deciding when to deinterlace as you need data from all 3 planes at once. in YUY2 you have all data at one place and can therefor be easier to work with. A plugin like Convolution3D that treats all planes in the same way must benefit enormously from YV12 though.

[trbarry]

TomsMoComp deinterlaces the 3 planes separately, and seems to get away with it. I seriously prefer working in planar format, though I could imagine situations where it would be nice to have more registers.

- Tom

[sh0dan]

Athlon64 will be a godsend for most assemblers - 16 general purpose registers + full and even extended SSE2.

However in my experiene there is rather seldom any problem in processing planes separately - even for deinterlacing. But I agree - for achieving "perfect" results (as decomb IMO) it requires to consider both chroma and luma. (But processing one or two lines at the time would probably be the most cache efficient).

[tangent]

Originally posted by primitive
YU12 actually contains less pixel data and is therefore fundamentally faster to work with.
Most filters tend to work only with the Y domain. Resize filters will actually be slower with YV12 because the chrome images need to be upsampled.

[trbarry]

Resize filters will actually be slower with YV12 because the chrome images need to be upsampled.

Chroma has to be resampled in YUY2 also, but you have to break it apart first and there is more of it. For my SimpleResize filter I didn't do any rigorous testing but I'm pretty sure the YV12 version is faster than in YUY2.

- Tom

[primitive]

YV12 is correct. Wow, I'm slow.

Please don't come down on me too hard for parroting information I saw elsewhere ;)

BTW, in the YV12 thread, most of the filter authors who were porting their filters to YV12 were reporting a 20-25% speed increase, which is consistent with the 6:8 ratio of pixel information relative to YV12:YUY2.

Don't most filters deal with chroma information? If so, why do they see such a speed increase?

(primitive)

[scmccarthy]

Those are reports of the speed increase of the total encode. Here if question 7 of the YV12 FAQ, bold is mine. What will be the main advantages?

- speed increase:
That depends entirely on the external plugins whether they will have YV12 support or not. Speed increases like 25-35 percent are expected. Of course there will only be a large speed increase if both your source and target are in YV12, for example in DVD to DivX/Xvid conversions.

- no color conversions:
The colour information doesn't get interpolated (so often) and thus stays more realistic. You see the speed increase is being credited to the lack of color conversion as well. And that improves the quality too, because if you interpolate to get YUY2 and then subsample again when encoding it to divx and then interpolate again while playing the divx file, that degrades the quality slightly.

Stephen

[benf2]

ok, then back to my original question of doing dvd to dvd using cce..should i use avs 2.5 or 2.07? which do u feel is best....i wish i was more up on this color conversion stuff. I always run avs without filters. Just basic load and source.:confused:

[scmccarthy]

My initial response was it depends on which filters you want to use, since there is a different set for 2.5 and 2.0x. There are a few other improvements in 2.5 that don't get discussed very often like the audio.

Stephen

[benf2]

with 2.5 i use either the mpeg2dec3 or mpegdecoder.dll

with 2.07 i use mpeg2dec2

[scmccarthy]

My advice is definitely stick to 2.07 for now. Or, if you use 2.5 use the mpeg2dec.dll which still serves up YUY2 and hasn't been changed yet. 2.5 is still only beta, the main fun of betas is figuring out how they don't work. On the other hand, there might be mmx optimizations that only apply to 2.5 and improvements to 2.0x are stalled. Bottom line is if you work in YUY2, the improvements are only marginal and no one is probably paying a lot of attention exactly how much.

Stephen