Best resolution to capture from VHS

[kefoo]

I'm a little confused by all of the literature on the resolution of VHS tape and the relationship between lines on a tape and pixels in an MPEG video. The general concensus is that VHS displays 240 lines, but that's not the same as a pixel. I want to capture video from VHS tape and convert it to MPEG-1/2 for computer viewing, while preserving as much of the quality and resolution of the source tape as possible, and keeping the aspect ratio straight (4:3) for the monitor to display it properly. File size isn't an issue, as long as the quality is good. I've tried capturing the same clips at several different resolutions and have found that I can get better quality video at reslutions higher than 352x240. How high can I go before I exceed the capabilities of VHS and start fabricating pixels? Should I go with non-interlaced MPEG-1 or interlaced MPEG-2?

Kevin

[Eric VanHeest]

If the temporary drive space isn't an issue at all, capture at as high of a resolution as possible. The higher the capture resolution, the more noise you can effectively eliminate when resizing to the final resolution. I don't know what size to use if your VHS source is "perfect" (which it can't be, since it's analog), but I've never seen an increase in resolution -not- help the final video, even up to the 704x480 (or whatever PAL is) range.

[kefoo]

I'm capturing at 704x480. Is the VCR actually putting out that many pixels, or is the capture card interpolating the signal to get to the capture resolution? I guess the question I should ask is "What is the best output resolution for the final video to best preserve the quality of the source tape?"

Kevin

[This message has been edited by kefoo (edited 04 September 2000).]

[Eric VanHeest]

I don't have any technical information about VHS to offer (I'm sure other people do), but when I reduce the final video to 320x240, it doesn't look as good when expanded back to full-screen as when I use 448x336 as my final resolution. I'm not sure whether 448x336 is any better than 384x288; I've been doing experiments with those two resolutions and they seem to be very similar at the same MPEG bitrates (this is after applying temporal cleaning, deinterlacing, and whatnot to the source video).

I'd have to say (completely subjectively) that 320x240 is not high enough to preserve the source quality. 384x288 might be ideal, and anything greater than 448x336 is almost definitely a waste of space. I used 400x300 for a while, but multiples of 16 seem to make certain programs happier, so I would stick to 384x288 or 448x336, depending on what you think is better. One thing I've definitely discovered through the process of learning about video is that dealing with anything analog is incredibly pesky!

Incidentally, the best thing I've found to use for random video experiments is video with a lot of text, such as subtitles or credits. It's fairly easy to see the difference between 320x240 and 448x336 video when watching text (expecially small text).

[paulw]

What sort of data rate are we talking about here. I usually use 1.9 or 2.5 Meg/sec @ 704 x 576 PAL.

[Dr Mordrid]

A 240 (NTSC) high capture (288 PAL) will never reproduce full quality because this only records one field of the two fields in a standard video frame. You can only get the full vertical resolution using 480 for NTSC or 576 for PAL.

As far as what horizontal resolution you should capture at that depends a lot on the nature of the footage and what you're going to do with it.

If it's "natural" video (scenery, talking heads, home movies etc.) then half width resolution works great (usually 352 or 360). I use it all the time and it has the added advantage of using half the disk space for a given time recorded. With natural video it's very difficult to see the difference between the half and full width settings.

On the other hand, for a scene that will have a keyed effect (bluescreening), titles or graphics added it's better to use the full resolution available on your card (usually 704 or 720 wide). This is to preserve as much of the detail as possible in the final product. In this situation the difference between full and half resolution can become very apparent in the final product.

Dr. Mordrid

[Grigory]

Hi,

The vertical resolution MUST match the number of active lines in video - 480 or 576 for NTSC/PAL. This and only this resolution keeps both fields in video signal and preserves full motion video editing and output to tape.

You cannot get better resolution than 352 pixels in width from VHS. It is theoretically impossible, IF the capture card works as expected. Most of them do.
The difference between 352 and 704 is visible on PC screen if you put some text on your video, but the result, recorded on VHS tape will be the same. So, if you wish to see better titles on PC screen, but get the same final video on tape, you can use 704 width. If you want to see ~ what you get - use 352 or 384 width.
Yes, editing wrong aspect ratio video may give you some problems, but at least Premiere allows to preview wideo in correct aspect ratio window. MJPEG video on TV will also look correct with 352 pixels width.

Actually, all ordinary TV sets with composite inputs with 14-21' in size require ~ 480 pixels width frames to produce as sharp as possible picture on screen directly from PC, not from VHS tape. So, even DV to MPEG encoding is sometimes better to do at 480x480(576) size, because of less file size with only minor loss in resolution.

So, capture from VHS with 352-384 (what is possible with your card) frame width, and exactly 480(576) heigth and enjoy fatser processing and less space on disk.

Grigory

[LvR]

For interest sake

Can somebody define the hight of a "line" (PAL) in terms of a pixel/dot?

And then define the gap between 2 "lines" in terms of a pixel/dot as well

Is it then possible to make up, using Photoshop/Coreldraw, a frame, that can be repeated and then recorded in an avi sequence to "easily" test the quality of connections/cabling? (Sort of something like the BBC testcards)

------------------
Lawrence

[This message has been edited by LvR (edited 05 September 2000).]

[Todd]

My understanding (and if I'm wrong, somebody, Doc? please correct me) is that the number of lines in video is the number of seperate vertical black bars that can be distinquished on a white background. That would make a line one pixel and the space one pixel, giving a pixel width of twice the line count. If this is correct then 400 lines would require 800 pixels.

There are test patterns available to determine video resolution in terms of lines. I've found some on the net, but I don't have the URL handy. If I find it I'll post a followup.

Todd

[This message has been edited by Todd (edited 05 September 2000).]

[Todd]

Here are some excellent references for analogue video resolution.

<a href="http://members.aol.com/ajaynejr/vidres.htm">http://members.aol.com/ajaynejr/vidres.htm</a>
<a href="http://www.elitevideo.com/new2.htm">http://www.elitevideo.com/new2.htm</a>
<a href="http://www.bealecorner.com/trv900/respat/index.html">http://www.bealecorner.com/trv900/respat/index.html</a>

That last one is a test of the trv900, but it also has links to the resolution test image (EIA 1956) in various file formats. These are intended to test cameras, but you might be able to use one of the lower res ones as a video source.

You can get other test patterns from <a href="http://www.displaymate.com/patterns.html">http://www.displaymate.com/patterns.html</a>

Todd

[LvR]

Thanks guys

Very usefull and interesting links

------------------
Lawrence

[Grigory]

TV resolution is the number of individual picture elements in the biggest square, cut from 4:3 TV screen.
For vertical resolution, we have 576 or 480 active lines, and this is the resolution. This resolution gives you the ability to see 288 black and 288 white lines, total 576.

For horisontal resolution, remember that it is measured in square, so for 720 pixels only ~540 are inside this square. This is the definition of horizontal resolution of DV video format.
For analog video, to measure the resolution, you have to apply 100% contrast sine wave to the input, and measure the frequency that gives 70% contrast on TV screen. Then count how many bright AND dark lines are inside the screen width, equal to the screen height, and this gives you technically correct resolution value. Knowing the screen aspect ratio and scan line duration you can easily calculate MHZ to lines.
The test patterns-based measurement is not 100% precize, because you have to tedect a point where the individual lines become invisible. This may happen at 50% or even 30% contrast, so the value is often little bit higher than for oscilloscopic measurements.

VHS has 240 lines, that gives for digital equivalent 352 pixels for frame width.
Laser disk has 544 pixels and ~400 lines
DV has 720 and 540 lines.

The reason why TV uses "lines" is simple - the number of horizontal scan lines is always fixed by standard. So, it might look convenient to count how many "lines" are necessary to reproduce all details in the input signal, if the scan lines would go vertically instead of horizontally. For VHS this value is 240*4/3=320 (NTSC) and 288*4/3=384 (PAL). 352 pixels for VCD and other half width video were taken as an average value.

Grigory

[Brian Ellis]

Just a little additional info. Whereas I agree with most of the foregoing, there is a little point that should be considered. If you push a very-high resolution signal into a VHS VCR, it is perfectly true that what is recorded on the tape will not have the same resolution. The VCR will have removed the highest frequency components from the signal, in effect, like a low-pass filter. If you push a lower-resolution signal into the VCR, it will still effectively pass through the low-pass filter and, because low-pass filters do not have a perfect step response (and if they did would be prone to ringing), you will lose a little bit more of the high-frequencies that remain. I don't think that the difference would result in a very noticeable visual change when viewing a tape on a TV from a normal viewing distance in your living room, but it is easily discernable with test patterns with resolution wedges when viewed close to.

My view is therefore to record a VHS tape from the best quality signal you can, to obtain the best possible quality, bearing in mind that I'm being more royalist than the king and that a single poorly-aligned, worn or dirty head would make even more difference!

------------------
Brian (the terrible)