Interlace MPEG2 Encoding

[Ejoc]

I was thinking about how interlaced video gets encoded for MPEG-2s. And there 2 ways I think it could be done, but im wondering which it is, or if its not what I was thinking at all.

So we know that for an interlaced frame you have 2 fields of the same width but 1/2 the height, so for an SVCD 480x240A and 480x240B. Does the encoder combine the 2 to make a 480x480, encode that, and then break it back up into 2 Fields, or does it encode each field as its own frame.

Im wondering because I find some interlaced DVDs to have really poor image quality, and I'm wondering why.

Or if I'm totally off, can you tell me or point me to a ref that explains how interlaced video gets encoded?

[DrWafer]

if you are going to make a VCD,SVCD or DVD out of your mpeg2 files
then you should leave it INTERLACED.

the only case that you need to deinterlace is when make a DivX or other
mpeg4 file wich you ll keep on your hard drive and play on your pc.

when deinterlace the encoder uses differend technic every time,bob technic combines the two fields into one frame,some others simply dismiss
every second field,odd or even.
no matter the technic the result is always 25 progressive frames.

but when leave it interlaced then the stand alone DVD player produces 50 frames out of 50 fields and the result on TV Screen is perfect on moving scenes.
otherwise when deinterlaced the DVD player produces 50 frames for TVset
but first and second,third and forth,fifth and sixth.....etc are identical and therefore motion is jerky.

expanation of theese here:
http://www.100fps.com/

[Ejoc]

Thanks, *But* that wasnt my question. I'm asking how the encoder encodes a frame when the output is interlaced. My question has nothing to do with "How To" or "Which should I use" Its a Technical question about how it gets encoded by the encoding program. Thats why there so few responses.

Option 1, the encoder basicly is encoding each field, without any ties to the other field.

or

Option 2, the encoder combines the two Fields, encodes, and seperates them back out.

Picture quality has to do with your bit rate, and changes from the frame before to the current frame. So if the fields are encoded seperately wouldnt you have a lower picture quality then if the the two fields were combined and the frame was encoded, then spilt back out into seperate fields.

Thinking out loud:
I dont think the encoder would be able to seperate an encoded frame into fields, so it most likely has to encode each field seperately. Which would mean that an interlaced MPEG-2 could never have as good of picture quality as a progressive MPEG-2.

[Ejoc]

Hmm I found a PDF that talked about it some. I guess if you are wondering you can check it out:
https://www.videohelp.com/forum/viewtopic.php?t=172241

and other:
http://www.bretl.com/mpeghtml/pixtypes.HTM

So I guess it can encode either way, and it almost sounds like interlaced is better for high motion, where progressive is better for stils/high detail...

[vhelp]

Hi Ejoc,

No, I didn't read those links you provided (above)

But, based in my expeirence w/ Interlaced source and encoding them, the
encoder will encode them as if they are a full frame, and NOT on a field
per field bases. If that were the case, the quality would be that much better
than full frame based, bebause there would be LESS pixels to work with,
and more quality as a result.

Also, a good tip of this fact and any other sort of comparison you would
care to decisfer, is to do the same source (if you can) and encode both
movies (one minutes worth) .. with one in Progressive, and the other in
Interlace.. given the source to be field based and full-frame based and
fps at 29.970 for both sources.

Given the above, you'll soon learn that the full-frame would be the winner
in quality, because there are no Interlaced frame (distorations) that are
being encoded, hence the more blockyness in Interlace source!!

-vhelp

[Ejoc]

Thanks vhelp,

Yea I've always done progressive ~200 CD's worth. Whats kinda neat is, it looks like it can actually do it either way, with advantages and disads for each.

I was mostly trying to figure out why the Stargate SG-1 DVD are such poor quality, I mean after I work on them and convert them to a progressive picture and remove the horrible tecline problems the DVDs have, I think my CVDs are better then the DVDs, which just seems wrong.

I guess I like to think out loud sometimes

[RoopeT]

Most interlaced DVDs contain material originally intended for TV. Therefore, they are shot with completely different type of cameras (cheaper) than material originally created for cinema release.

When releasing big budget films on DVD, the original film is scanned at resolutions much greater than 720x480/576, I think this is the main reason for the quality difference.

[vhelp]

Hi Ejoc and others..

w/ respect to Interlace source material encoding...

Another thing to consider is the Encoder app (or hardware for that matter)
How well does it handle Interlace, and/or what technique does it use to
handle them.

Another thought, is the "users" skill level. How much does the user know,
in dealing w/ Interlace source materials ??
Does he/she know tricks/techniques of the trade (or not) or something other.
Or, how about his/her sklll level just with the SW app itself. There may,
hint hint, be tips/techniques to use in a given SW app when dealing with
Interlace source materails.

Yet, another thought.. what about the "Source" itself, and in combo w/ the
above, how well does the "user" deal w/ (knowing the source) Interace
encoding.

And another.., what about the "means" of obtaining the source, ie, it could
be DVD rip based; Analog capture based, DV capture based, already encoded
based, etc, etc.

Interlace comes in may different flavors. Yes, hard to believe, but thats
actually true. ie, frame based, field based, inter-mixed based, yada yada..
I've tackeled many different types.

Examples:
* DV based, from home footage
* Captured types..
***: diferenciate between the Station airing the source. Source could be
.....using a different form of Interlace (though, still Interlace)

I know the above may sound confusing, but I've run up against many of these
and delt w/ pretty much each and every one of them. There are all various
kinds of Interlace, and you have to know how they are Interlaced (type, that
is) and deal with them, based on type. Confusing.. hahh ?? Yeah. But, that
explaines (at least to me) why some Interlace materials come out great in
one encode, but bad in others, even when using the same "de-Interlace"
technique (or not) !!

Point.. did you know that DV Interlace is (or can be) different then from
a known source to be Interlace that you captured ??

nother point.. did you know that (as in case for DV cams) one CAM can
have a better Interlace system than the other, and hence, can be better
handled with a "de-Interlace" or "Interlace" technique ??

Some more stuff to consider, when you are at a stump.

-vhelp

[Kedas]

Hi something isn't very clear to me.

Is interlacing/de-interlacing not done during capture.
If you capture to AVI and encode to Mpeg2 later to put on DVD.
isn't the most important step the capturering and not the encoding ??

An other question if I capture 25fps interlaced isn't that really 50fps (half resolution)?

And if I capture 25fps progressive then that is 25fps (full resolution)

I want to put some tapes on DVD so I should capture iterlaced and then encode interlaced. That should give me the best quality right? because that is how it's present on the tape.

[FOO]

The main reason to encode interlaced is that
2 of every three frames from NTSC Film have fields
that were originally from different Film Frames.

Encoding a Frame composed of shots from different times means
that each line is unrelated to its neighbor. MPEG encoding relies
on nearby pixels to be similar for its compression effectiveness.

[pijetro]

Interesting question Ejoc. At first, I would have guessed that the encoder scans every second line, much like a TV set scans every second line, but I think that I would have been way off.

I'm starting to lean towards the encoder, creating both fields as a single frame. Here's the reason.

When I used Restream to change the field order, the flags were changed for even field, odd field. The only way that this could be done in reasonable time without reencoding, is if the original frame was encoded as a whole 480 line frame....
I myself encode purely interlaced sources produced by VHS tapes..As far as your quality issues are concerned, I'm wondering if there's something else in the way of your quality issues??

[FOO]

If you guys are trying to figure out how data is stored....
I have a copy of the ISO 13818-2 MPEG2 Spec.
Each picture (frame) can be stored as either 1 frame or
2 separate fields. Top and bottom fields can be flipped without
actually swapping the fields.

[Ejoc]

Yea I found that MPEG-2 interlace can be encoded 2 ways, each field by its self, or by combining, encoding and then seperating.

[Cornucopia]

Hi something isn't very clear to me.

Is interlacing/de-interlacing not done during capture.
If you capture to AVI and encode to Mpeg2 later to put on DVD.
isn't the most important step the capturering and not the encoding ??

An other question if I capture 25fps interlaced isn't that really 50fps (half resolution)?

And if I capture 25fps progressive then that is 25fps (full resolution)

I want to put some tapes on DVD so I should capture iterlaced and then encode interlaced. That should give me the best quality right? because that is how it's present on the tape.

This might clear some things up:
EVERY consumer and prosumer capture device, and the great majority of pro devices, are video-based and are interlaced.
The few exceptions here are digitally created files (3dstdmax,maya,aftereffects, etc renders) and the highest-end sony/rank/etc telecine/datacine machines (which are switchable) and the weird-market stuff like medical imaging.
So capturing will determine if the avi (etc) is interlaced.
However, if the source is NOT interlaced, it is before or during the encoding to MPEG that a decision should be made to de-interlace, or ivtc, or not. For example, if i'm shooting a billboard with a video camera on a tripod, even though the camera and digitizing is interlaced, the original source (the billboard) could be considered progressive. If it gets de-interlaced via simple recombining of fields (obvious no-motion required here) then resolution will be full frame instead of half frame, with no jaggies.

To help with your confusion, use clearer abbreviations:

Frames per second = FPS or Fps
Fields per second = fps

so 25 FPS = 50 fps !

If your original source is recorded via std. video camera (and NOT a screener) and you capture normally, and you intend to view with std DVD player on std TV, then keep the whole chain in interlaced for best quality. Otherwise, it depends.

Scott