Is 704x480 with a 16:9 DAR valid for DVD?

[poisondeathray]

Thanks, jagabo

Do you have any MPEG2 samples that have sequence_display_extension() present ?

If sequence_display_extension() is not present then it is intended that the entire reconstructed
frame is intended to be mapped to the entire active region of the display. The sample aspect ratio
may be calculated as follows:

SAR = DAR × horizontal_size / vertical_ size

[jagabo]

Do you have any MPEG2 samples that have sequence_display_extension() present ?

No. Or more properly, I don't know of any tools that report it, so I don't know if I have any or not.

[poisondeathray]

I think elecard streameye might be able to

[jagabo]

I don't have $1800 to spare!

[Alex_ander]

GSpot shows it, have seen 540x480 in 'seq ext' box for files with pan & scan. Most often, that box is empty. DVDRemake usually fills extension bits with default (actual) pixel numbers.

[jagabo]

GSpot shows it

Thanks, I usually use MediaInfo these days so I never noticed GSpot's seq ext field.

[intracube]

Found it:
https://forum.videohelp.com/threads/299570-Correct-size-Resolution-for-AVI?p=1989388&vi...=1#post1989388

See page 47: http://neuron2.net/library/mpeg2/iso13818-2.pdf

Thanks, jagabo. I'm sure I've read bits of that pdf some time ago. Time to revisit it.

I don't know of any tools that report it, so I don't know if I have any or not.

I'm going to look at the headers of a few files. Time to fire up the hex editor and get my hands dirty.

[autodidact]

I thought I would add that it seems that ReStream is able to detect if this Sequence Extension information is present in an MPEG stream and also add or remove it from the stream.

[Anonymous344]

Shon3i reports that 16:9 is valid for 704x480 in the DVD specification.

[edDV]

Back to the question of proper downsize from 1920x1080 to DVD. I wasn't part of those discussions back then but we can role play the DVD authoring pioneers. There are two choices, scale 1920x1080 to 704x480/576 or to 720x480/576.

1920x1080 to 704x480/576
This would maintain correct aspect ratio when 16x9 is letterboxed or converted to anamorphic. This works for broadcasters because they transmit 704 and analog conversion blanking crops to the equivalent of 704 (702 PAL).

1920x1080 to 720x480/576
DVD movie authors have mostly wide source usually 2.35 to 1 aspect ratio. In the early 90's there were D2 Laserdisc transfer masters optimized for letterbox inside native 4:3 NTSC or PAL. These mapped directly to 704x480/576.

So when they regeared for anamorphic wide later in the 90's, why did they extend out to 720*? My guess is 720 at the same PAR results in slightly thinner letterbox bars at the expense of 2.3% side crop when converted to analog or 704 digital. The full width is there if you view with zero overscan 720 digital. This can be done without aspect ratio distortion for 2.35 to 1 source.

This wouldn't be true for 4x3 (1.33 to 1) or native 16x9 (1.78 to 1) source since there is no letterbox to absorb the vertical size increase. If native 4x3 or 16x9 source is scaled to 720, aspect ratio is affected. The resulting picture is vertically squished. These should be scaled to 704 or 720 with 8 pixel left and right side bars.

Any other explanations?


*720 originated in the CCIR-601 (D1) standard when it was discovered repeated A/D conversion caused H shifts resulting in side clipping of the 704x480/576 frames. 8 pixels of side padding were added to protect the sides of the 704x480 image from clipping. At the transmission point, the picture was centered, then cropped to 704.

[Anonymous344]

*720 originated in the CCIR-601 (D1) standard when it was discovered repeated A/D conversion caused H shifts resulting in side clipping of the 704x480/576 frames. 8 pixels of side padding were added to protect the sides of the 704x480 image from clipping. At the transmission point, the picture was centered, then cropped to 704.

May I ask if you have a source document for this? (I believe you: I just would like a source to help me convince others.)

[poisondeathray]

Shon3i reports that 16:9 is valid for 704x480 in the DVD specification.

Thanks for reporting back & confirmation

[Anonymous344]

Thanks for reporting back & confirmation

You're welcome. Thanks for your help too.

The fact that Restream can read these Sequence Extension information is invaluable. It should make it possible to see how HC Enc writes the Sequence Extension information.

[edDV]

*720 originated in the CCIR-601 (D1) standard when it was discovered repeated A/D conversion caused H shifts resulting in side clipping of the 704x480/576 frames. 8 pixels of side padding were added to protect the sides of the 704x480 image from clipping. At the transmission point, the picture was centered, then cropped to 704.

May I ask if you have a source document for this? (I believe you: I just would like a source to help me convince others.)

This was a hot topic during the mid '80s tests leading to CCIR-601 (later ITU-Rec601). The result in the spec was provision for 720x480/576 at the same PAR as 704x480/576. When MPeg took up ITU-Rec601 as the basis for MPeg2, they also offered both frame sizes and maintained backward compatibility for broadcast applications but for optical printing, 720 was often favored since the guard pixels weren't needed. Many ignored the same PAR issue but the DVD standard maintains 704 and 720 at the same PAR for compatibility with broadcast.

The "at the same PAR" issue implies 4x3 and 16x9 are defined as 704x480 with 720 extra pixels extending the width. The 2.35 :1 DVD movie authors took liberty with movie resolution as a creative decision thus they could say cropped 704 was the movie with 720 adding extra width.


This famous summary from the broadcast side contains most of the key references.
http://lipas.uwasa.fi/~f76998/video/conversion/

[hank315]

704x480 is a valid MPEG2 resolution for DVD (4:3 and 16:9).

In VMGM_MAT at byte offset 256, there are 2 bytes which show the video attributes.
Bit 2 and 3 in byte 1 hold the valid DVD resolution:

value    NTSC           PAL
0       720x480        720x576
1       704x480        704x576
2       352x480        352x576
3       352x240        352x288

[Anonymous344]

@ hank315

May I ask how HCenc writes the sequence_display_extension information for 720x480? Does it write it as if the image is contained in the core 704x480 image area?

[hank315]

The hor/ver resolution is in the sequence header, max. 12 bits for horizontal and 12 bits for vertical.
So 704x480 is simply written as hor=704, ver=480 etc. HCenc will handle all values as long as they are mod 16.
In the sequence extension header, these can be expanded by 2 bits.
But the MPEG2 standard limits the resolution to 1920x1152 (1995 standard), this is changed to 1920x1088 in the 2000 standard.
So the values for horizontal and vertical in the sequence extension header are always zero.

@jagabo
For some nice (free) tools, look here:
http://tsviatko.jongov.com/index.htm

[jagabo]

HcGui doesn't add the sequence_display_extension field.

[jagabo]

The hor/ver resolution is in the sequence header, max. 12 bits for horizontal and 12 bits for vertical.

That is not what we've been discussing. Just about every tool that displays information about MPG files will report those values. We're talking about the optional sequence_display_extension field which includes information about what portion of the frame represents the 4:3 or 16:9 image.

[hank315]

AFAIK, there's no extra info in the sequence_display_extension field, it's only an extension by 2 bits for horizontal/vertical (2 most significant bits), there's no relation with 4:3 or 16:9.

[jagabo]

AFAIK, there's no extra info in the sequence_display_extension field, it's only an extension by 2 bits for horizontal/vertical (2 most significant bits), there's no relation with 4:3 or 16:9.

See page 33* of the h.262 document I linked to earlier:

http://neuron2.net/library/mpeg2/iso13818-2.pdf

It defines an optional field called sequence_display_extension with an additional 14 bits each of width and height data. Pages 56 gives details about their usage.

*As numbered on the pages themselves, not the PDF page selector.

[intracube]

AFAIK, there's no extra info in the sequence_display_extension field

There's some useful tables here:
http://dvd.sourceforge.net/dvdinfo/mpeghdrs.html#ext

I think elecard streameye might be able to

The two options you've highlighted on your screenshot belong to sequence_extension. We're talking about sequence_display_extension and the associated display_horizontal_size, display_vertical_size.

Elecard Streameye might still display the relevant data (if the disc has it). It's possible the disc you tried didn't have it.

I've manually looked at the headers of some DVDs - a few do have sequence_display_extension, but most don't. Also, if I'm reading the hex values correctly, there's some weird values on some of the discs...

[Cornucopia]

I concur with edDV, and so do the Spruce Maestro, Sonic Scenarist, DVD Studio Pro, DVD Architect manuals, Andy Parsons of Pioneer (I have a few of his whitepapers), segments from DVD Forum's Books (that I can get my hands on without paying $5-15k) and if that isn't enough thoroughbred for you, there's no pleasing you!

Scot

[poisondeathray]

The two options you've highlighted on your screenshot belong to sequence_extension. We're talking about sequence_display_extension and the associated display_horizontal_size, display_vertical_size.

Yes you're correct

I've manually looked at the headers of some DVDs - a few do have sequence_display_extension, but most don't. Also, if I'm reading the hex values correctly, there's some weird values on some of the discs...

can you upload a sample that has sequence_display_extension() specified

EDIT: Nevermind, I dug around and found some.... But if you can explain what you mean with the weird value ones... ?

[jagabo]

I've examined a handful of DVD rips I have on my computer right now with GSpot. Some had no sequence_display_extension, some said 720x480, and one said 540x480.

The last one may not have been a commercial DVD. It's a video that should play 16:9. MPCHC and KMPlayer (both using their internal MPEG 2 splitters and decoders) play it 4:3 (squished horizontally), VLC plays it 16:9. GraphStudio (using Microsoft's mpg2splt.ax, and ffdshow's MPEG 2 decoder) plays it 16:9. GSpot says it's 4:3, MediaInfo says 2.35:1.

Those results where when playing the VOB files directly. When playing from the DVD menu (IFO) they all play 16:9.

[edDV]

I've been thinking about this 1920x1080 downsize to DVD issue. By that I mean full frame 16x9 source not 2.35 to 1 movies. If you downsize to anamorphic 704x480, the aspect ratio would be correct but those that view as non-overscan 720x480 would see the 8 pixel left and right sidebars. The vast majority would watch on an overscanned monitor and see 704x480 or less before monitor upscale.

So how could one properly downsize for full 720 picture width? I think the answer is to vertically crop the HD frame so that it fits 720x480 (or 720x576) at correct aspect ratio. Here is my first attempt.

720/704 = 1.023 horizontal expansion.

1080/1.023 = 1056.44 source vertical crop .

1056/16 = 66.0 ... i.e divisible by 16.

One would vertically crop 1920x1080 to 1920x1056 before downsize to 720x480 or 720x576. The center 704x480 or 704x576 would represent 16x9 at approximately correct aspect ratio.

[Alex_ander]

So how could one properly downsize for full 720 picture width? I think the answer is to vertically crop the HD frame so that it fits 720x480 (or 720x576) at correct aspect ratio.

'Correct AR' by mpeg2 ITU's standard H.262 clearly means 'to fill a standard 4:3 ок 16:9 display' with the part (or whole) of the stored image, defined in mpeg2 file: a) either by all sample numbers (like 720x480) in absence of S.D. extension*, or b) by specified number of those samples - if that extension exists and reads different numbers (like 540x480 for pan & scan, or smth. like ITU601' numbers as given in H.262 example)**.

* (p.47)
If sequence_display_extension() is not present then it is intended that the entire reconstructed
frame is intended to be mapped to the entire active region of the display.

**(p.176)
Sample aspect ratio information is provided by means of aspect_ratio_information [3 versions of DAR + 1:1 SAR] and (optional) display_horizontal_size and display_vertical_size in the sequence_display_extension(). Examples of appropriate values for signals sampled in accordance with Recommendation ITU-R BT. 601 are given in Table D-1.

(the table with an example, p.177)
display_horizontal_size / display_vertical_size

525-line 711 483
625-line 702 575

We have not seen here any serious document, saying the numbers derived from ITU601 give the only correct version of SAR (=PAR) for DVD. DVD is storage standard (not intended to be necessarily recorded exactly as captured), ITU601 is digital broadcasting / program exchange standard between studios (as its title says).

So if an encoder doesn't force ITU601 numbers in sequence_display_extension, there is no reason to adjust for 720/704 difference, just resizing anything known as 16:9 to 720x480(576) must be correct for any device with a standard mpeg2 decoder built-in.

[intracube]

EDIT: Nevermind, I dug around and found some.... But if you can explain what you mean with the weird value ones... ?

One of the discs - from House Season 4 (region2) has the following sequence_display_extension header:
00 01 B5 23 05 05 05 08 72 12 00
In this instance, the relevant bytes (8-11, in bold) convert to binary:
00001000011100100001001000000000
00001000011100 1 00001001000000 000 (reformatted for clarity)
001000011100 = 540
001001000000 = 576
The disc is 16:9 and uses the full 720x576. If the iso13818-2 doc is right, there shouldn't be any need for the sequence_display_extension at all (the DVD player should default to mapping the 720x576 region directly to a viewable 16:9 image. The horizontal value being 540 is odd - and would result in the central part of the image being stretched out over a 16:9 display and would look very distorted.
The disc plays fine in a standalone player, which makes me think these values are being ignored (or I've got them wrong).

Another disc 'The World Is Not Enough' has a sequence_display_extension; and the associated video_format as '000' (component) rather than '001' PAL - but it has the correct 720x576 as the active region.

[hank315]

Your binary math is fine, 08 72 12 00 is 540x576.
If the hor disp value is 540, it might be that pan/scan is active.
Does the stream also have picture display extensions?

[chowmein]

What the **** are you all talking about???