Another Ghosting Issue/Frame Blending Dillema

[jagabo]

The player (or TV) will stretch the full frame to match the size of the screen. If the frame has black letterbox bars you will black letterbox bars on the screen.

DAR = PAR * FAR

DAR = Display Aspect Ratio (final shape of the picture)
PAR = Pixel Aspect Ratio (the shape if individual pixels)
FAR = Frame Aspect Ratio (the frame dimensions)

16:9 DAR NTSC 720x480 DVD has a PAR of 32:27. 16:9 DAR PAL 720x576 DVD has a PAR of 64:45. You must change the PAR to convert from one to the other. Ie, you must change the shape of the individual pixels.

NTSC:
DAR = PAR * FAR
16:9 = 32:27 * 720:480
16/9 = 32/27 * 720/480
1.778 = 1.185 * 1.5
1.778 = 1.778

PAL:
DAR = PAR * FAR
16:9 = 64:45 * 720:576
16/9 = 64/45 * 720/576
1.778 = 1.422 * 1.25
1.778 = 1.778

[hizzy7]

I am starting to learn some stuff. I am watching videos on pixel aspect ratios. The shape of the pixels can vary, depending if it is PAL or NTSC, if I am understanding correctly. NTSC pixels are square, but PAL pixels are more rectangular. Is there a numeric value of the standard pixel size? If I understand, for PAL would it be 1.0925 for 4:3 For NTSC, it would it be 0.90 in 4:3, but those sizes would be different if they were widescreen.

Also, if I understand, resolution does not effect picture size.

Thanks!

[jagabo]

No DVD uses square pixels. You can calculate the PAR for 4:3 DVD using the equations I gave earlier:

PAR = DAR / FAR

NTSC:
PAR = 4:3 / 720:480
PAR = (4/3) / (720/480)
PAR = 1.333 / 1.5
PAR = 0.889
PAR = 8:9

PAL:
PAR = 4:3 / 720:576
PAR = (4/3) / (720/576)
PAR = 1.333 / 1.25
PAR = 1.066
PAR = 16:15

[hizzy7]

I follow right up until the last step. How do you get PAR = 16:15 or PAR 8:9?

Thank you!

[jagabo]

1.0667 = 16/15
0.889 = 8/9

Or you could perform the math purely as fractions:

NTSC:
PAR = 4:3 / 720:480
PAR = (4/3) / (720/480)
PAR = 4 * 480 / 3 / 720
PAR = 1920 / 2160 (then divide each by 240)
PAR = 8 / 9

[poisondeathray]

nm...jagabo beat me

[hizzy7]

Do people really call it 0.889 or is it often called 0.9?

Also, I when reading online I have seem mention of these stats:

PAL 720x576 standard 4:3 (1.0940)

Would the numbers in parenthesis signify the PAR? If so, shouldn't the PAR be 1.066?

Thank you again! You guys should start a school!

[jagabo]

If you do the same calculation with 704x480, 4:3 DAR you get a PAR of 0.909.

[hizzy7]

Its this video I have been watching to further understand PAR. Its from this guy that I read that the PAR of 720x480 NTSC video is 0.9

http://www.youtube.com/watch?v=ke9QzHGhF6w

Also, I added this edit:

Also, I when reading online I have seem mention of these stats:

PAL 720x576 standard 4:3 (1.0940)

Would the numbers in parenthesis signify the PAR? If so, shouldn't the PAR be 1.066?

Thanks!

[jagabo]

The DVD spec refers to the MPEG 2 spec regarding aspect ratios. The MPEG 2 spec is very clear. There are only a handful of display aspect ratios supported by MPEG 2 (the PAR is never specified, except in the case of square pixels where the PAR is 1:1) , and of those only two are supported by DVD: 16:9 and 4:3 DAR. The MPEG 2 spec is also very clear that the entire frame comprises the display aspect ratio*. So the full 720x480 or 720x576 frame should be stretched/squished to 16:9 or 4:3. So the pixel aspect ratios for DVD are the ones I gave earlier. This is at odds with the ITU 601 spec universally used for digitizing standard definition analog video sources where the inner ~704x480 or ~702x576 portion of the frame comprises the 16:9 or 4:3 DAR, even if the frame is padded with borders to bring the width up to 720 pixels. If you subtitute 704x480 or 702x576 into the equations I gave you'll see he ITU spec results in the PAR values you listed. Those are correct for ITU captures but not correct for DVD.

In reality, nobody in the industry cares about the difference between the two specs. Many DVDs made from video tape sources (like some old TV shows) retain the ITU PAR and are technically incorrectly encoded for DVD. Film sources tend to follow the MPEG 2 spec. DVD players are often schizophrenic with respect to aspect ratios too. Most often the upscaled digital output follows the MPEG 2 spec, the analog composite and s-video output follows the ITU spec.


* There is a mechanism by which only a portion of the frame can comprise the indicated DAR -- the use of a Sequence Display Extension. I've only ever seen that used on commercial DVDs to indicate a 4:3 portion of the frame with 16:9 videos -- for pan-and-scan purposes. And that value has always been 540x480 or 540x576. That matches the MPEG 2 PARs I gave earlier.

[2Bdecided]

jagabo has given a perfect summary.
(Though I wouldn't have called the MPEG-2 spec "clear" - it's oversimplified because the authors didn't think about what they were doing - they didn't sit down and intend to break compatibility with every ITU Rec. 601 video converter in the world!).

There really is no need to go into this level of worry, or even think of the PAR, for typical processing though. For progressive ~24p NTSC>PAL conversion you just resize 720x480 to 720x576, and make sure the 4x3 or 16x9 flag on the final MPEG encode is set to match that on the source. (and fix the video frame rate and resample the audio of course!).

Forget about the PAR. You don't need to know it.

Cheers,
David.

[hizzy7]

Hi Friends!

I am back with another query for the video gurus. I have a clip here which I have a question about:

http://files.videohelp.com/u/183506/sword.demuxed.m2v

There's all kind of mess going on here. Is it just interlacing artifacts or a bad NTSC/PAL conversion? What script would be best? Something with QTGMC? Tdecimate?

Thank you again for your help!

h

[jagabo]

It looks like the usual PAL to NTSC conversion but interlaced chroma was treated as progressive chroma so the colors of successive fields have blended together.

[hizzy7]

It looks like the usual PAL to NTSC conversion but interlaced chroma was treated as progressive chroma so the colors of successive fields have blended together.

Hi Jagabo!

So, I would have to more than just:

QTGMC(Preset="placebo")
SRestore()

to fix it, or is it not fixable?

Thanks!

[jagabo]

I don't think the blended chroma is fixable.

[hizzy7]

Is blended chroma a weird kind of color smearing? I just want to make sure I am noticing the right (or should I say, wrong) thing.

Thank you!

[jagabo]

Is blended chroma a weird kind of color smearing?

Yes. An interlaced frame of video has two fields, each with it's own colors. When mishandled the colors of those two fields get blended together.