What "happens" with a capture like this?

[absinthecarolinas]

I'm capturing some old VHS camera tapes using passthrough on my DV cam and WinDV. WinDV has always captured my DV tapes directly from the camera at 720x480, and I assume that's actually the resolution at which the camera is filming.

But these home videos are also being captured at 720x480, so my guess is that the camera always outputs that resolution. I mean, I don't know much about the older hardware, but I'm talking about VHS tapes filmed with a VHS home video camera circa 1990. I sort of doubt it filmed at 720 x 480, but I don't know!

The point I'm getting at is: Am I correct in thinking that these particular 720x480 captures through the DV cam weren't natively that resolution? And does it therefore follow that trying to reencode them (to DVD) without deinterlacing would be the wrong thing to do?

thx,

-abs

[edDV]

You can sample them at 720x480 fine. The VHS source has a maximum native resolution around 352x480 interlaced. A VHS source will have both fields for the full 480 lines. Simple deinterlacing will convert this to 240 lines, tossing half the existing vertical resolution. Why do that?

With DV pass-through, you will have a 720x480 DV file. You can encode this to 720, 704, 640, or 352 x480 for a playable DVD. I'd experiment with 720 and 352 and see if you see a difference, then make your decision.

Edit a sample clip with representative scenes and use that. Evaluate the results on the TV from a DVD playing in the DVD player. A computer monitor will distort the results.

[SLK001]

640 x 480 is not a valid DVD resolution (but the others are). A VHS tape has an equivalent resolution of 352 x 240.

The VHS camera recorded the video in analog format, so there is no "native" resolution.

Since most video cameras film fields sequentially (then combine two for a frame), do NOT deinterlace the video.

[reboot]

I always thought that one needed to cap VHS in at LEAST 640x480, preferrably 720x480.
"The VHS video works with the interlaced system (not progressive), which means it draws every frame twice on the screen. In fact, every frame of an interlaced video is divided in two fields. Those two fields mean that every frame carries much more than just 240 or 288 lines of information. If you capture at 352x240 or 352x288 you'll be skipping every second field, which results in only half of the resolution."

Is this hogwash?

[absinthecarolinas]

Man, this sh*t can be confusing

I'm thinking like reboot. I thought interlaced video was created by capturing two fields, at the same moment in time, and interlacing them. In other words, it's almost like it is made up of progressive images ... each image just has artifacts of interlacing in the form of those little lines that can be seen if motion scenes are slowed down.

So when I ask, "should I deinterlace?" I simply mean use an AviSynth filter like FieldDeinterlace to "blend" the lines together. That doesn't throw away resolution, does it?

Or, if I'm wrong (hey, it happens) ...

* So, if I understand this correctly, NTSC vertical resolution is 480 in essentially all formats ... DVD, VCD, SVCD, and VHS (and TV??). So all that is happening when I do a "passthrough capture," at least in terms of resolution, is that the analog signal is converted to 720 horizontal 'pixels.'

I'm just trying to figure out here if there's any significant quality loss in this capture process.

I should have mentioned that I had been planning to convert these clips to XviD, my codec of choice. But I have very little experience doing interlaced XviD encoding, and in fact I'm not sure it isn't a touch buggy. I have been able to get it to play back in my Philips standalone, though. Apparently, XviD defaults to bottom-field-first.

-abs

[edDV]

640 x 480 is not a valid DVD resolution (but the others are). A VHS tape has an equivalent resolution of 352 x 240.

The VHS camera recorded the video in analog format, so there is no "native" resolution.

True, 640x480 is a valid ATSC DTV spec but not DVD and equivalent is a better term.

The analog spec for NTSC VHS puts maximum luminance bandwidth at approx 3.0MHz.. Nyquist theory states that analog sources must be sampled at at least 2x bandwidth so the minimum sample frequency should be 6.0MHz. This would correspond to 313 horizontal samples minimum without producing alias products. Closest DVD spec is 352.

Since most video cameras film fields sequentially (then combine two for a frame), do NOT deinterlace the video.

Fields are never combined into a frame until they reach the eye and brain.

I always thought that one needed to cap VHS in at LEAST 640x480, preferrably 720x480.

That varies by the capture technology of the capture board itself but either 640x480 or 720x480 exceed nyquist's rate. As stated above 640 isn't a DVD supported sample rate so 720 is best. Many brooktree cards compromise by capturing to 640x480 and then resample to 720 or 352.

A DV camcorder capture (inside the camera) is always 720x480 which is what you are using. The DV camera has the additional advantage of better NTSC decoding than the typical capture card.

"The VHS video works with the interlaced system (not progressive), which means it draws every frame twice on the screen. In fact, every frame of an interlaced video is divided in two fields. Those two fields mean that every frame carries much more than just 240 or 288 lines of information. If you capture at 352x240 or 352x288 you'll be skipping every second field, which results in only half of the resolution."

Is this hogwash?

It is close but not specific.

"In fact, every frame of an interlaced video is divided in two fields." add each separated in time by 1/59.94 sec

"Those two fields mean that every frame carries much more than just 240 or 288 lines of information. If you capture at 352x240 or 352x288 you'll be skipping every second field, which results in only half of the resolution." ... should end with which results in only half of the vertical resolution.

[edDV]

...

I thought interlaced video was created by capturing two fields, at the same moment in time, and interlacing them.

No, standard video has fields displaced in time by 1/59.94 sec and that is what makes deinterlacing so difficult. Simple deinterlacers toss a field (and half the vertical resolution) to avoid having to deal with motion between fields.

24fps film is handled differently, I'll add a link below.
...

...
So when I ask, "should I deinterlace?" I simply mean use an AviSynth filter like FieldDeinterlace to "blend" the lines together. That doesn't throw away resolution, does it?

Blend will blur the fields together. If there is motion, blend will produce smeared blurs in the motion area. Fancy deinterlacers use motion adaptive algorithms.

...
* So, if I understand this correctly, NTSC vertical resolution is 480 in essentially all formats ... DVD, VCD, SVCD, and VHS (and TV??). So all that is happening when I do a "passthrough capture," at least in terms of resolution, is that the analog signal is converted to 720 horizontal 'pixels.'

VHS is NTSC. The capture device must also decode the NTSC to YUV. Luminance is sampled into 720 pixels yes. U and V (I and Q in reality) are sampled into 720/4 = 180 pixels per line in a scheme called 4:1:1. PAL uses 4:2:0 sampling of YUV.

The "vertical sampling" has already been done.

...
I should have mentioned that I had been planning to convert these clips to XviD, my codec of choice. But I have very little experience doing interlaced XviD encoding, and in fact I'm not sure it isn't a touch buggy. I have been able to get it to play back in my Philips standalone, though. Apparently, XviD defaults to bottom-field-first.

If your goal is to play XviD on a computer display directly, then deinterlace is necessary. If you do this, you will have a substandard playback from a DVD player to a TV.

Alternative is to keep it interlaced (full vertical resolution) and use a deinterlacing player (e.g. PowerDVD) when viewing on the computer. It will do a better job deinterlacing than a simple blend. Likewise, a good DVD player will do a superior job deinterlacing for a progressive TV display.

Added> read this to understand film and progressive DVD players.
http://www.dvdfile.com/news/special_report/production_a_z/3_2_pulldown.htm

[reboot]

Thankyou for explaining that EdDV.
I just know, that capturing VHS at anything less than full D1 aspect, results in crappy video (on my system/software), so I now either cap in virtualdub for avi, or windvr 3 for mpeg-2, both at 720x480.

[absinthecarolinas]

Thanks for the explanation, edDV, though the bit about colorspaces goes over my head. I haven't read up on the differences in YUV, YUY, RGB, etc. and frankly don't have much of an understanding of it.

But if I'm wanting to preserve these clips, captured from VHS, by encoding to XviD files for watching on TV ... would you agree that it sounds best to encode them as interlaced? Seems like a standalone DivX player would play back an interlaced XviD just as it would anything else interlaced. They also look quite tolerable viewed on a computer screen (of course, this is 15-year-old VHS, so the standard isn't real high, ifyaknowhatimean).

I definitely need a better understanding of colorspaces. So far, my biggest complaint with the interlaced XviD encodes I've done from these captured samples are color differences between the original captured RGB avi and the XviD output.

I'm going to try encoding a few to DVD MPG for comparison and see how that looks.

-abs

[edDV]

...
But if I'm wanting to preserve these clips, captured from VHS, by encoding to XviD files for watching on TV ... would you agree that it sounds best to encode them as interlaced? Seems like a standalone DivX player would play back an interlaced XviD just as it would anything else interlaced. They also look quite tolerable viewed on a computer screen (of course, this is 15-year-old VHS, so the standard isn't real high, ifyaknowhatimean).

My comments mainly apply to MPeg1&2 for standard DVD.

I haven't played much with the XviD/Divx players but it seems they are very sensitive to bitrate and other encoding specs. They offer a small decoding window due to limited dsp power in the early players. You need to carefully match encoding to what they can play.

Future generation players will handle MPeg, Divx, XviD, Wmv (RC-1) and h.264 all the way up to HDTV spec. and will have enough horsepower for decoding.

I definitely need a better understanding of colorspaces. So far, my biggest complaint with the interlaced XviD encodes I've done from these captured samples are color differences between the original captured RGB avi and the XviD output.

Computer monitors are RGB, progressive (deinterlaced) with 0-255 somewhat linear grayscale, bright phosphors and fast refresh rates.

TV sets are Y,U,V* (aka Y,Pb,Pr analog or Y,Cb,Cr digital), mostly interlaced**, 16-235 grayscale, non-linear gamma, darker phosphors and slower refresh rates. A very different animal.

* Y is lumiance or the monochrome part of the image. UV are the color difference part of the image.

** NTSC and PAL systems are interlace only. The current DVD standard calls for interlace disc encoding, but both fields can be present for progressive reconstruction (see the link above).

New digital broadcasting standards can be interlace or progressive, standard definition or high definition. Next generation HD DVD will handle any of MPeg2, RC1 (wmv), h.264 (MPeg4 variants).