PAN-SCAN ?!?!

[AndreMetello]

Hi

This is probably a dumb question, but anyway....

I was watching an interview with a dvd-auhorer on tv these days and he said "today we can author a movie into both widescreen and full screen format, using the same video file, because of pan-scan technology thats avaiableo n most dvd players.."... Whats exatcly this pan-scan feature and how could one prepare a video for that ?


thanks a lot

[SaSi]

Pan and Scan is a feature of the MPEG-2 video standard since 1995, so this guy was telling yesterday's news.

Now, since you are a newbie, you are excused but don't ask questions about Pan And Scan here. It's taboo. People get excited because pan and scan displays only a portion of a video.

What Pan & Scan is about is encoding, within the video stream, information about the center of interest of the movie, so that a window is created in the movie to display only the part that "is interesting" or actually, can fit on the display screen. That's why the information encoded is called "Display Extensions"

How this information can be inserted is an interesting question that I've asked before and never got an answer. In the meantime, I've found out a few more things.

1. It must be inserted in the MPEG-2 video stream before it's authored.
2. In theory this can be automatically generated by the Encoder (while it is busy doing the motion search for the compression).
3. Again, it can be manually entered after the encoding by a special application.
4. Again, in theory, it is possible for the MPEG-2 encoder to generate crude Pan And Scan codes, insert it in the video stream and then use a specialized application (viewer with editing capability for the P&S fields) to fine tune this manually.

In practice, the only relevant thing I've found is an option in CCE to include P&S in the stream. I've not tried it yet to find out if it does insert information or just blank extensions.

In any case, for those wanting to be able to view a 16:9 movie in a 4:3 frame TV, P&S is much better than hard croping the edges - or even worse cropping the edges slightly and stretching the vertical axis to fill the screen.

[adam]

Yet strangely enough, almost no commercial DVDs implement this feature. That's why we have dual layered disks half full with WS half full with FS, dual sided DVDs with WS on one side and FS on the other, and completely separate DVD releases one in WS and one in FS.

From what I have read, this is one of those things considered a FEATURE in the DVD standard, not a requirement. So apparantly not all DVD players support this option, and thus its not widely used for compatibility purposes. Its a real shame cause it could have solved the whole P&S vs WS debate before it even got started.

[SaSi]

BTW Adam, do you know if checking the Pan Scan checkbox in CCE will actually include P&S vectors ?

[vhelp]

Hmm..

Hi guys.

This is all interesting Way back, oh saying about 2 1/2 years ago, when
I got my first DVD-ROM drive (still have it) I remember playing some DVD
movies from it, to compare w/ my AD-500, anyways.. and I remember
seeing a PAN&SCAN feature in the software (SW) player that, when clicked
would allow me to use my mouse as a "hand", and I could drag the movies
scene left/right.

Now, given the above, am I correct to assume that this was a P&S feature
for that DVD movie ??

I can't remember which movie.. was either:
* Total Recal; Contact; Moonraker; Alien; Dogma ..

or one of those movies. I thought it was interesting, but at the same time,
I said, "why would anybody want to scroll the movie all the way to the right
on their PC monitor for ?" Anyways..

That was my expeirence w/ P&S in my first step days hehe
-vhelp

[SaSi]

vhelp,

Try PowerDVD with a DVD. Go to full screen and then right-click and in the Pan And Scan entry select an aspect ratio. It fills the screen and allows you to "manually" pan and scan. The idea behind P&S feature is to have the decoder automatically know how to scroll the "display window" within the movie.

[adam]

No idea Sasi. I've never done much more than play with any versions of CCE past 2.5, and I don't believe 2.5 has this option.

[FOO]

I have been reading ISO 13818-2 lately and remember seeing
where the vectors are. It would seem that if one of the encoders
could insert fixed "center screen" vectors, it would be just as
good as having both WS and FS in the same stream.

I wonder why they don't do that ?
Can CCE insert them from some file maybe ?

[SaSi]

I have been reading ISO 13818-2 lately and remember seeing
where the vectors are. It would seem that if one of the encoders
could insert fixed "center screen" vectors, it would be just as
good as having both WS and FS in the same stream.

I wonder why they don't do that ?
Can CCE insert them from some file maybe ?

Inserting fixed Center Screen vectors will simply crop the movie sideways. However, it will, as you may think, cause the video to become full frame.

As was mentioned by someone before, the P&S feature is MPEG-2 syntax that any decoder MUST understand, however the player may simply ignore it. So, it appears that some DVD players are not able to do P&S display.

Mine, for example, has a setting in configuration where I can select Letterbox, wide screen or P&S, so I can assume that the player will "obey" the P&S vectors if any exist.

Now, even if your favourite MPEG-2 encoder cannot insert display extensions, you can. Their presense in no way gets in the way of the normal video stream. Actually, according to the MPEG-2 specs, you don't need to add too many Display extensions. I suppose one is enough, as supposedly the player must use the last P&S vector information it has.

You can write a small program that inserts a picture display extension after the first picture header extension. For fixed vectors centering the display window, the display extension will be fixed. You can insert it within the normal MPEG-2 video stream.

Perhaps, (I'm not sure) you need to do that after every sequence header if the occurence of the Sequence header resets the P&S vectors, but it shouldn't.

The complete way to do it would be the following:

Parse the video stream, and decode down to the motion vectors for each macroblock. For every frame, make an array of 90 elements (for the 720 horizontal pixels) and in each element store the total value of motion vectors for each vertical slice of macroblocks. Calculate the Gaussian distribution of these samples and use the center of the curve as the display center. Generate Picture display extensions driving the center of the screen left or right, depending on the max. of motion vectors.

The approach above does the following in simple terms: It calculates the most "active" part of the picture frame using the motion vectors as a measure and scrolls the center of the display area to match the "center of action", if it can be called that way.

The important detail is to use motion smoothing, i.e. not allow more than 2-3 pixels movement per GOP - for example - otherwise the display will jitter left and right like crazy. Like someone using a Camcorder and panning left and right. You won't be able to view the footage.

I'm working on an MPEG-2 decoder program, not to decode and playback video but to explore the MPEG-2 stream and make objective (mathematical) calculations of image quality. I've reached the stage where I've decoded the macroblocks and generated the quantized array. IDCT is a big jump in terms of code but I realize that the code has reached the stage where it can perform the above scenario and then re-write the MPEG-2 video stream adding Picture display extensions.

I may give this a try.