some simple VCR -> DVD capture/processing questions

[ricnews]

I want to transfer some VHS tapes to DVD, such that they look decent viewed on TV or PC.

I've got a CompPro VideoMate TV Gold + (I think) and can get decent decent mpeg2 captures using Compro's PVR software (which I believe is based on Intervideo WinDVR). Actually, I can only get decent capture up to around 6400 kbps, as above that motion gets quite jerky. I don't really understand why that is since the CPU load is < 100% and the hard drive should be fast enough. Any idea?

After capturing a crop the noise at the bottom then resize the video back to original size and recode in TMPGEnc 3.0 or in Virtualdub frameserving to TMPGEnc 2.5. The results are pretty bad, with lots of blockiness not present in the original capture.

Is this due to the resizing, or re-encoding mpeg2 or both? If I just mask the noise at the bottom with black pixels rather than resizing the cropped video, will this prevent the mpeg2 capture from becoming further degraded?

At the moment I'm trying to capture to mjpeg using VirtualVCR, which I'm going to do while I sleep. I'm guessing this is a better approach since there isn't so much loss in the original capture. The downside is it uses much more disk space. OK for now, though.

With this approach, will I be okay to resize the video, to avoid having the black bar at the bottom?

The other question I have, is do I need to deinterlace the video before resizing? If I'm not planning to use any additional filters, is there any advantage to deinterlacing at all?

Thanks!
Ric

[The Geek]

Is this due to the resizing, or re-encoding mpeg2 or both?

Both.

If I just mask the noise at the bottom with black pixels rather than resizing the cropped video, will this prevent the mpeg2 capture from becoming further degraded?

No. But the degradation won't be as much as with the resizing.

At the moment I'm trying to capture to mjpeg using VirtualVCR, which I'm going to do while I sleep. I'm guessing this is a better approach since there isn't so much loss in the original capture. The downside is it uses much more disk space. OK for now, though.

That is the correct way to do this. Yes, it takes alot of space, but well, that is necessary in order to get the highest quality possible. With hard drives so cheap nowadays, you can get a dedicated capture hard drive. It doesn't need to be big. A 60 GB drive would be enough.

The other question I have, is do I need to deinterlace the video before resizing? If I'm not planning to use any additional filters, is there any advantage to deinterlacing at all?

Well, resizing an interlaced video will degrade the quality, too, but not much. At least the results is not blockiness, it's just a bit softer than the original.
But deinterlacing is the worst you could do. The quality will suffer alot.

The best way to do what you want is this:
1) Capture in a format without much quality loss (like MJPEG for example)
2) Load it into TMPGEnc, resize it to remove the blurriness at the buttom
3) Let TMPGEnc output to MPEG-2, using two-pass VBR and a high bitrate of about 7000 average, 8000 maximum. Set the quality setting to "Highest quality".

The Geek

[Arachnotron]

The other question I have, is do I need to deinterlace the video before resizing? If I'm not planning to use any additional filters, is there any advantage to deinterlacing at all?


Never resize an interlaced source vertically! In doing so you mix the two fields resulting in a blurry picture no amount of filtering will ever get right again.

Is this due to the resizing, or re-encoding mpeg2 or both? If I just mask the noise at the bottom with black pixels rather than resizing the cropped video, will this prevent the mpeg2 capture from becoming further degraded?

Resizing. Just mask and you will be ok.

[ricnews]

In terms of vertically resizing an interlaced video, what about using the folding (or unfolding?) filter in virtualdub to separate the fields, resize, recombine the fields, then crop the bottom? Shouldn't this prevent the blurring of the 2 fields?

I was happy to do the mjpeg capture, until I played it back and found the audio was way out of sync. At the very beginning it was fine, but within a few minutes it was a second or more off and getting worst. I'm thinking of picking up a Hauppauge WinTV-PVR250 with hardware mpeg2 encoding. I'm thinking this would allow me to capture in mpeg2 at 12 mbps or higher if it allows, to minimize quality loss (because it will have to be re-encoding after masking/resizing to a lower bitrate to fit on the DVD anyway) but not have to worry about dropped frames and audio sync as hopefully the hardware will handle all that appropriately.

Is high bitrate mpeg2 comparable to mjpeg for this purpose? If I'm going to capture in a non-mpeg format there's no point in buying the Hauppauge (although it'll still make a better PVR...).

In terms of the audio sync problem with mjpeg and virtualvcr, are there any guides that help in preventing this? I've tried and failed many times to re-sync so I'm convinced if I can't get the sync right during the capture, I'll never get it right. I also tried capturing in virtualdub, but for some reason, the capture works only in 160x120 and 320x240 modes. Any other resolution returns just a solid green screen. I couldn't figure out why.

Thanks again,
Ric

[Arachnotron]

In terms of vertically resizing an interlaced video, what about using the folding (or unfolding?) filter in virtualdub to separate the fields, resize, recombine the fields, then crop the bottom? Shouldn't this prevent the blurring of the 2 fields?

IIRK the internal resize in vdub has a switch for interlaced contents. (and the option to letterbox it for you too)

In general, unfolding the fields, rersizing and refolding is ok if your resizer cannot handle interlaced stuff.

VirtualVCR has some audio resampling options which might help you.
Look
here (http://www.doom9.org/capture/capturing_VirtualVCR.html) for more info. (there is more usefull stuff in the capture guide by the way about processing interlaced stuff)

The vdub problem has to do with vdub being a vfw program and your capture drivers being WDM. It is a known problem, and there have been various postings about it. There is also some stuff in the various faqs.

[edit] before resizing, first check if it is really needed and what it will do to your aspect ratio.

Again, read the analog capture guide for more on this stuff.