Multicam SD + HD interlacing(?) problems

Jay-Hancock wrote on 6/27/2006, 1:38 PM
I have a multicam project where one cam is an HDV and the other is an SD camera. I render it to a CF DI. On playback, the portions where the SD footage is active look really jaggy whenever some motion is happening. It's like the person moving has a pinstriped shirt (and face and hands!).

In the project, I've specified (in media properties) that the SD footage is lower field first and the HDV footage is upper field first. Overall project and render properties specifies that the output is upper field first.

Is this jaggy video problem happening because I mixed upper/lower field footage? The HDV cam footage comes out pristine in the render.

Comments

Spot|DSE wrote on 6/27/2006, 3:20 PM
Which are your project settings? HDV or SD?
Are you delivering to HD or SD? If you are delivering to SD, you shouldn't be worrying about UFF/LFF. Just set up your project for a DV output, and let Vegas downconvert the HD to SD for you without worry.
Marco. wrote on 6/27/2006, 3:28 PM
Also be sure you have one of the Deinterlace Methods selected in the Project Properties. Having "none" selected may cause problems like the one you described when scaling is involved.

Marco
Jay-Hancock wrote on 6/27/2006, 3:30 PM
Project settings is HDV. Final outputs are both high def and SD. I am creating a CFDI which will then be encoded to different distribution formats. I'll render it to an HDV file for my own use, and I also render it to a PAL DV avi for import to TMPGENC and ultimate output to a VCD for a European user.

Doing the full render in vegas to CF DI took 23 hours, which I prefer to do only once. From there I wish to encode to final forms (SD PAL and HDV), which I found only takes a couple hours.
Jay-Hancock wrote on 6/27/2006, 3:32 PM
Marco, that does confuse me, since the final product is interlaced (not progressive). Why would a deinterlace method need to be specified? I don't doubt you, I just wonder what's the logic/reasoning behind it.
Marco. wrote on 6/27/2006, 3:40 PM
Selecting a Deinterlace Method in the Project Properties does not do deinterlace your output. It only changes the way signals are handled. You can't do any harm to have "Blend Fields" (or even "Interpolate Fields" selected there. This also is the standard preference.

Deinterlacing only will be applied when you also choose progressive render settings in addition to certain Project Properties.

After you said you use HDV project settings I'm almost sure the Deinterlace Method is the cause for your problem if also you had chosen "None" here before.

Try setting the Deinterlace Methode to "Blend Fields" in the Project Properties and I bet a coin your problem is gone.

Marco
farss wrote on 6/27/2006, 4:05 PM
Also don't forget to render at Best.
Jay-Hancock wrote on 6/27/2006, 4:59 PM
Thanks everyone for helping me out here. I was using best all along (which is probably why it takes 23 hours to render a 20 minute video).

Marco you hit the nail on the head. In military traditions, if someone pulls out a unit coin it means you either pull out a better coin or buy the person a drink. I guess I should owe you both, a coin and a drink!

My guess is that Vegas internally must combine the fields for its own purposes, so that it has a whole frame to work on.

I found that even without the jaggy interlace problems, faces in the SD footage get blurred to almost oblivion in the high def scaled output. So I needed to add a bit of sharpening. I used the unsharp mask for this. And then I used Mike Crash's smart smoother (after sharpening) to reduce the resulting noise. Probably there are better ways to accomplish this, but fortunately the SD cam's footage is only being used for a few short parts of the overall project.

So many things to learn...
fldave wrote on 6/27/2006, 7:30 PM
There was a thread a long time ago about upsizing video, with the best method was using VirtualDub. I tried it and it looked pretty good. You might try resizing your small clips of SD and see if that works better.

Basically unfolding the fields, resizing using Lancos, then refold. From Laurence:

http://www.sonymediasoftware.com/forums/ShowMessage.asp?ForumID=4&MessageID=354140

I use the native filters that come with latest version of version of VirtualDub version 1.6.

You need to use a chain of three filters:

1: deinterlace filter set to "unfold fields side by side"
2: resize filter set to lancos 3 resize and resizing to a number that is half as high and twice as wide as what you want to end up with.
3: deinterlace filter set to "fold side by side fields together"


Jay-Hancock wrote on 6/27/2006, 8:24 PM
I think I'll try this sometime soon. Thanks for the tip, Dave!
Marco. wrote on 6/28/2006, 12:34 AM
We'll have a beer together anywhen anywhere. ;-))

Marco