Truth about DVD & Compressed Video

Jay Gladwell wrote on 6/19/2004, 9:49 AM
It was not my intention to high-jack anyone's thread, but I read something that Catwell posted in the thread Burning to DVD and I wanted to increase the odds of getting an answer by posting it here.

Catwell said:
A point I learn in yesterday's VASST seminar. . . If you have media on the Vegas timeline all transistions, titles, or generated media are uncompressed. If you render it to DV you are then compressing it. If you then compress that file to MPEG 2 you are compressing a compressed file.

I do not have the definitive answer to that, but it doesn't sound accurate to me. I could be wrong. Have been before. One or twice. ;o)

Can anyone else substantiate this, one way or the other?

I was under the impression that DV was compressed from the moment it was recorded. Then, in the render, Vegas just "copied" the 1s and 0s as they are in the timeline to a "new" DV file.

J--

Comments

taliesin wrote on 6/19/2004, 10:00 AM
What he wrote there is not false but I don't understand what he is aiming for. Of course titles (if made inside Vegas) and Generated Media (if made inside Vegas) is not yet compressed before rendering. Transitions? - Transitions are not media till it is not rendered so you cannot compare transitions with titles or Generated Media.
Now when you render titles or Generated Media to dv of course it will be compressed. But using that compressed dv files to render to MPEG then this is not a copression of compressed files but the dv files will be decoded/decompressed first.
To keep best quality it should be better to use an unrendered timeline for MPEG encoding inside Vegas (because this saves from doing a coding/decoding-generation) - though the loss you'll get if you render to dv first is not that very much noticable.

Marco
Nathan_Shane wrote on 6/19/2004, 10:04 AM
Well...I do recall Spot saying here in Dallas that the Media Generators in Vegas produce "pure uncompressed digtal" at 4:4:4

So I think perhaps Catwell was referring "only" to that media which Vegas produces itself, not media that is brought in from outside of Vegas.
Orcatek wrote on 6/19/2004, 10:41 AM
My understanding was that once vegas applies a transition etc, the result is uncompressed. This is how it is dealt with by the application. This prevents and recompression problems as fx and other items are layer with this resultant video. It is only compressed when it is output.

Makes sense as the intermediate format is clean this way.

JohnnyRoy wrote on 6/19/2004, 11:47 AM
What Spot was referring to was the fact that anything Vegas modifies, whether its generated media, transitions, or color corrected footage, Vegas operates on uncompressed using a 4:4:4 color resolution. DV avi uses 4:1:1 color resolution. MPEG2 uses 4:2:2 color resolution. So by rendering to DV and then to MPEG2 (instead of MPEG straight from the timeline) you are going from 4:4:4 to 4:1:1 to 4:2:2. Obviously you have already lost a lot of color information which you won’t loose of you render as MPEG2 directly from the timeline which takes the internal 4:4:4 down to 4:2:2. This will not affect DV media that is untouched but it will make your titles and other generated media look better in MPEG2. So it wasn’t so much the compression as it was the color resolution.

~jr
Catwell wrote on 6/19/2004, 2:47 PM
JohnnyRoy expresses exactly what I was trying to communicate. Your MPEG 2 file will have more accurate information if it is generaqted from 4:4:4 media than from DV media. Even if you decode the DV you do not have the color data that was lost in the encoding. I probably should not have used the word compressed as it is not the exact problem. The problem is the loss of information when you encode in any lossy format. You cannot get the information back.
busterkeaton wrote on 6/19/2004, 2:51 PM
So

4:4:4 to 4:2:2

is better than

4:4:4 to 4:1:1 to 4:2:2
Jay Gladwell wrote on 6/19/2004, 2:56 PM
So just to make certain I understand, we're only talking about "generated media" created in Vegas, not the DV video recorded by the camera, correct? How can "color corrected" DV footage be 4:4:4?

J--
JohnnyRoy wrote on 6/19/2004, 3:02 PM
We are definitely talking about the generated media but also any DV media that has been altered. So if you added any video FX like color correction, that color correction is being done at 4:4:4 internally. That means that any processed media has the potential to look better. Since the original DV media was 4:1:1, I don’t know how much better it will look but I used color correction specifically because it has the potential to add colors that may not have been in the original. It’s like having 16bit audio files be processed internally at 24bits to add FX and then dithered back down to 16bit. You will get a better sound than just applying the FX at 16bit internally too Straight DV media that was not touched will remain at 4:1:1 and will not have any improvement.

~jr
John_Cline wrote on 6/19/2004, 3:07 PM
For what it's worth, NTSC DVD's are not 4:2:2, they are 4:2:0. If you make a 4:2:0 MPEG2 file from a 4:1:1 DV .AVI file, you end up with essentially 4:1:0.

John
JohnnyRoy wrote on 6/19/2004, 3:13 PM
Yea, I thought Spot said that DVD MPEG2 was 4:2:0 but I wasn’t sure. When I looked up MPEG2, it said it was 4:2:2 but I do remember Spot saying 4:2:0. Is this just DVD MPEG2 or all MPEG2?

~jr
farss wrote on 6/19/2004, 3:30 PM
As far as I know ALL mpeg-2 is 4:2:0, probably converts better from PAL than NTSC as PAL is 4:2:0 to start with. This is part of the reason the best mpeg-2 encoding is obtained using hardware encoders that go straight from 4:2:2 (or better) material via SDI.
Unless the material is being played out on top line DVD players and fed component into a studio monitor I seriously doubt you'll see much of a difference. Probably 90% of the DVD players are connected via composite to a TV in which case I wouldn't loose too much sleep over it.
It'd be simple enough to do some tests, use Vegas to produce a video using only generated media, preferably something with various rates of motion and colors etc. Render straight to mpeg-2 and render to avi and render that to mpeg-2. Burn DVD with the two videos and watch. Nothing beats a real world test.
discdude wrote on 6/19/2004, 3:37 PM
I was under the impression that Vegas processes everything (transitions, generated media, etc) in the RGB colorspace, not YUV 4:4:4. While the two are functionally equivalent in terms of color resolution, they are not identical.

In theory, YUV 4:2:0 (or 4:1:1) to YUV 4:4:4 upsampling is faster than a YUV to RGB conversion.

In any case, both DV and MPEG-2 are lossy codecs and it is generally advisable to avoid recompression as much as possible. I would render straight to MPEG-2 unless I had a better codec than the Mainconcept codec included with Vegas.
Jay Gladwell wrote on 6/19/2004, 3:43 PM
Well, all this is obviously over my head. I guess because I'm looking at it as if it were an image, say in Photoshop. If I have an image that was created at 72 ppi, taking that image and resizing it to 300 ppi and reducing it to 144 ppi isn't going to make it any better. But let me ask this question:

What does each space in the "4:2:0" for example, represent? What is each digit referencing?

J--
discdude wrote on 6/19/2004, 3:59 PM
Y = Luminance
U = Blue - Y
V = Red - Y

The first number position refers to Y, the second U and the third V.

The numbers refer to the sampling rate. Essentially the frame is broken down into groups of 4. 4 means the color information of every pixel is recorded. 2 means only half the info is stored. 1 only a quarter of the info is stored.

Hope this helps.
taliesin wrote on 6/19/2004, 4:17 PM
>> As far as I know ALL mpeg-2 is 4:2:0

No. Definetely not.

Marco
B_JM wrote on 6/19/2004, 7:01 PM
there is a 4:2:2 profile for mpeg -- but its not supported for dvd before anyone asks ..

riredale wrote on 6/19/2004, 8:45 PM
You can find out more about what things like 4:2:2 mean at the first few sites on this Google page.

DV in NTSC countries is 4:1:1, which has great vertical color resolution but lousy horizontal color resolution. DV in PAL countries is 4:2:0, which is more evenly-balanced in my view. DVD MPEG2 is also 4:2:0, as mentioned.
Spot|DSE wrote on 6/19/2004, 9:05 PM
Videocurmudgeon,
I walked into this late, sorry...(delayed flight from Seattle this a.m.)
Video is PROCESSED at 4:4:4, but nothing in the end, unless you render to uncompressed, is 4:4:4.
The point of that discussion is, if you are printing to tape and printing to DVD, you want to render to BOTH avi and mpeg as separate renders for the optimal result.
If you render from your Vegas timeline to DV and THEN render to MPEG, you suffer some quality loss. You'll notice it more in some media than others. Gradients can really demonstrate this.
So, what I showed was that you can use the Batch script to render to both. The amount of time to render to avi and to mpeg as separate, independent files isn't significantly longer than rendering to avi and then rendering to mpeg in all situations. (sometimes it can be damn slow though, depending on timeline content)
If you render to an avi only, you'll get a fast render to mpeg from that avi, but you'll also throw away another chunk of the overall sample. For low motion/no motion media containing few motion titles and gradient, or detail like smoke, moving water, etc, you may not notice the loss. But if it contains grades, moving titles, water, flame, smoke, mist, blurs...you'll usually notice.
Grazie wrote on 6/19/2004, 9:11 PM
So, Spot, is there an argument for rendering "difficult" parts as UNcompressed and take advantage of the superior results? Creating a series of Copmpressed & Uncompressed files and route them off to DVDA?

I'm trying . .

Grazie

rs170a wrote on 6/20/2004, 5:29 AM
"What does each space in the "4:2:0" for example, represent?"

http://www.adamwilt.com/DV-FAQ-tech.html#colorSampling will give you the technical details of the sample rates and
http://www.adamwilt.com/pix-sampling.html has a number of images to graphically show you.
BTW, this is an excellent website that youcould literally spend hours browsing around. HTH.

Mike
Jay Gladwell wrote on 6/20/2004, 7:10 AM
Thanks, Spot, that I can understand. I've never claimed to the be the sharpest knife in the drawer, so, as usual, I appreciate you're taking the time to help out with these things.

Jay
farss wrote on 6/20/2004, 7:48 AM
SPOT, while I'd absolutely have to agree with you on generated media I cannot see how if you've shot in DV and applied no FXs how it's going to make any difference if you render to DV and then mpeg-2 or go straight to mpeg-2. Either way you've already taken the hit of DV compression and Vegas is only making a bit copy.
What may also have a bearing and so far hasn't been discussed is the effects of DV compression as well as color sampling.
Chienworks wrote on 6/20/2004, 8:40 AM
"applied no FXs"

That's the critical point. If you're rendering unaltered DV material then there is no benefit to avoiding a DV render. However, anywhere that you do have any alterations, effects, crossfades, fades, filters ... etc. (anywhere that Vegas will need to render) then you will lose more going through a DV render to MPEG than straight to MPEG. Keep in mind that in these areas, Vegas is generating new media out of your existing media.
Spot|DSE wrote on 6/20/2004, 8:48 AM
Grazie,
I can't think of any time you'd be rendering for uncompressed to go straight to DVD.
If the source footage was DV to start with, the only time the information in this thread applies is when you are working with edited media.
When you color correct, when you title, when you insert images, when you create composites, when you use masks, when you create generated media, when you use TM or PC, or do ANYTHING to that media, it's processed as uncompressed. This is why you see the "frame recompressed" indicator in your preview window.
Next...
When you print to tape, the media MUST be DV if you are using Vegas' print tools. So, you'll render to DVin an avi format for this print.
If you are rendering to MPG for DVD delivery, you want the most information to go into that render. So, you render to mpeg from the Vegas timeline too. If you render from your timeline to avi and then render that avi to MPEG in DVDA, you take a quality hit. See John's post and my post above, or read what Kelly just wrote. Between the 3 of us, I hope this makes sense.
For you being in PAL land of DV being 4:2:0, you shouldn't see a hit, but frankly, I've not experimented with that. I need to before we leave on the Oz trip, I guess.
Anyway to sum up, except for archiving or sharing media between applications, there is very rarely reason for rendering to uncompressed.