How many lines is NTSC overcan?

[cym]

I capture a lot of Hi8/8mm/VHS tapes. I'm capturing these to a PC in DV format at 720x480 over firewire. Of course, all have the overscan lines at the bottom. However, I also see a lot of tapes with quite a bit of noise (varying static, snow) at the bottom, perhaps 15-20 lines. The final destination is typically DVD (sometimes the AVI file for hard disk is also requested). When a DVD is played on a TV, the overscan is hidden, and often much if not all of the noise at the bottom. But, when a DVD is viewed on a PC, or if the AVI is played on a PC, the noise is plainly visible. I'd like to clean it up for the PC viewing.

So, my questions are:
1) How much of the 720x480 captured area can I mask, before I am masking into the visible region when played on a TV?
2) Is there a preferred approach to mask symmetrically (top/bottom/sides)? Or just the noise at the bottom?
3) Is there anything that can be done to reduce/eliminate the noise, or is it inherent in the recording from a cheap camera, or tape aging, etc? (FWIW, my heads are clean)

Thanks.
CYM

[jagabo]

The amount of overscan varies from TV to TV. With CRTs it varies with temperature, age, brightness of the picture etc. Just mask the noise and don't worry about whether you can see the black bars.

[cym]

Thanks, but there must be some minimum per the NTSC specification. All captures I've seen have around 8 lines of overscan at the bottom, which I've never seen on a television, implying the minimum is at least 8 lines. I don't have the NTSC spec, but I thought someone who has read it know how many lines must be outside the visible region, per the spec.

[jagabo]

Thanks, but there must be some minimum per the NTSC specification. All captures I've seen have around 8 lines of overscan at the bottom, which I've never seen on a television, implying the minimum is at least 8 lines. I don't have the NTSC spec, but I thought someone who has read it know how many lines must be outside the visible region, per the spec.

My guess is that most CRT TV manufacturers shot for about 5 percent overscan at each edge (ie, the "action safe" area that broadcasters use). But it doesn't really matter. Any individual TV varied from that the minute it left the assembly line.

In my experience, most fixed pixel HDTVs (LCD, plasma, etc) overscan by 2 to 3 percent at each edge. But I've heard of more and less. And of course, many also have a non-overscan mode where the entire frame is visible.

[solarfox]

Thanks, but there must be some minimum per the NTSC specification.

Actually, no, there's not. The NTSC spec does not define anything beyond the details of the actual broadcast signal (frequencies, voltage levels, color and sound subcarriers, etc.); it doesn't impose any particular requirements on how the signal is decoded and displayed.

[MOVIEGEEK]

Overscan is done by the TV manufacturers, it varies from company to company and model to model.
I have seen it vary from 2-15%.

Try Enosoft DV Processor.

[Steve Stepoway]

The NTSC spec does define the Vertical Blanking Interval which does not contain "active video", but is part of the 525 lines recorded in an analog recording. Some are used for synchronization (defined in the spec), and some for closed captioning and time code (not defined in the original spec, but perhaps in a revised version -- that VITC is how VCRs set their clocks -- and part of what you sometimes see in a mal-adjusted analog TV, or at least you could in the olde analogue broadcasting days). But all those should be stripped away in digitizing to a 720x480 image. I'm sure edDV can recite the details a whole lot better than I can.

Steve

[edDV]

The lines of noise you see at the bottom are due to head switching on analog VCR's (VHS, 8mm, Hi8). Digital recording has no noise at the bottom.

As said above there is nothing in the NTSC or ATSC spec that specifies minimum overscan. I've done a quick search of SMPTE standards and found nothing for overscan. The Movie and TV production communities usually define "Action Safe" as 5% each side or roughly the center 670x432 of the image. But action safe is inside typical overscan.

The BBC has redefined action and title space much tighter for digital transmission. For a 4:3 picture they put action safe at 3.5% top and bottom. This implies overscan is expected to be less than 3.5% on a digital TV.
http://www.bbc.co.uk/guidelines/dq/pdf/tv/tv_standards_london.pdf P-8, P-19 to 21

[AlanHK]

1) How much of the 720x480 captured area can I mask, before I am masking into the visible region when played on a TV?

Just mask out the static; if your mask is visible it'll be a lot less annoying to have plain black than flickering static.

[cym]

edDV: Thanks for the spec on 3.5% you mentioned below. That at least gives me some guideline. Should I read it to mean 3.5% total, or 3.5% at the top and the bottom. I say that because the diagram on page 20 in the spec shows 10 lines reserved at the top for this, but it also implies this is an interlaced diagram, so that means 20 lines. The total screen size in the diagram is 574 and 20/574 = 3.48%. Is that how you interpret it?

AlanHK: I think we can do better than just mask it. It may be better to mask the bottom and recenter, without resizing, as an example. But, one needs to know where the center is of the masked image, and how much one can mask to place the picture correctly, and still fill the viewable screen. That's what I'm trying to understand here.

[edDV]

Cym, I'm still not sure what you want since overscan depends on the display. Many displays can be set for no overscan (e.g. Samsung "Justscan"). In that case, any masking will be visible.

At the other extreme, a failing high voltage power supply for a CRT TV would have variable overscan depending on average field luminance. This is called "blooming".

If you are creating video for others to enjoy, you would follow action and title safe rules for average viewers but also provide quality video to the edges for no overscan viewers. If you mask, those people will complain*.

If you are doing this for yourself for a particular TV set, then measure the overscan for that TV. The next TV you buy will be different. Most likely a new digital TV will have less overscan.


* most won't complain about mild letterbox but they might complain about side bars for 16:9 HD.

[2Bdecided]

AlanHK: I think we can do better than just mask it. It may be better to mask the bottom and recenter, without resizing, as an example. But, one needs to know where the center is of the masked image, and how much one can mask to place the picture correctly, and still fill the viewable screen. That's what I'm trying to understand here.

That doesn't make sense. If you're going to mask the junk, and re-centre it without resizing, you don't need to know anything about the display device.

Whether you see the mask (or not) depends on the amount of overscan in the display device, and the amount of junk you need to mask. You don't have any control over either of these things!

btw, when you re-centre it, beware of breaking interlacing (i.e. keep the move to mod2), and try to avoid resampling YV12 chroma (i.e. maybe keep the move to mod4 or mod8 if working with YV12).

Cheers,
David.

P.S. While I often mask the junk and centre what's left (just as you have suggested), be aware that this (obviously!) means shifting the image within the frame - and so any carefully framed shots may be less optimal, or at least not quite what was intended. Not likely to be noticed.

[jagabo]

And with MPEG family encoding you want to keep the borders mod8 or mod16 in size to prevent DCT ringing at the edges.

[cym]

2Bdecided: Your right. My comment about knowing where to center didn't make sense. But also, your point about the effect of recentering, carefully framed or not, indeed may not work out well. Even someone's home movie will have the subject's in the middle, and if I have to remove 50 lines from the bottom and recenter, even that may not look right. So far I haven't recentered anyone's video, just masked the bottom for cases where they view on a PC or as has been mentioned here on TVs that can display the full screen. Thanks for the point on the effect on interlacing on shifting it. I hadn't considered that.

jagabo: I wasn't aware of the effect on MPEG. I understand how that would affect the DCT (discrete cosine transform), but hadn't considered that aspect of MPEG encoding.

[jagabo]

jagabo: I wasn't aware of the effect on MPEG. I understand how that would affect the DCT (discrete cosine transform), but hadn't considered that aspect of MPEG encoding.

It also has an effect on motion vectors (which work on 8x8 blocks). You'll a little loss of efficiency if you don't have mod8 borders.

[2Bdecided]

It makes less difference than you might expect - and zero difference if you're encoding interlaced content "properly" (i.e. because it shouldn't really have hard horizontal edges!) ...

http://forum.doom9.org/showthread.php?p=1257932#post1257932

Cheers,
David.

[jagabo]

I didn't expect a large difference. I've run similar tests myself. And the nature of the video make a difference too. Vertical pans will show a bigger difference than horizontal pans, for example. Having a sharp edge between the picture and the black border fall on a non-mod8 boundary will also lead to more DCT ringing artifacts near the border.