from 4:3 DV to 14.4 (pseudo 16:9) DV

[dennismv]

EDIT: this method may be good for those with original 4:3 footage who want to convert it to play in 16:9 format but without cutting off too many lines of resolution.
14.4:9 removes 75 lines of resolution and has black bars on the sides that extend when played as as 16:9. (Converting to true 16:9 without black bars means removing 120 lines of resolution, which I found to be too much.)

Method:
What I did (I will put resolution first for nicer formatting):
720(.9)x480 -- original standard DV footage (.9 is pixel aspect ratio (PAR))
720(.9)x405 (14.4:9) ignoring PAR, I blacked out (letterboxed) 37 pixels from top and 38 pixels from bottom, thus converting it to pseudo-16:9, as 720:405=16:9.
[To clarify: Why 14.4:9 ? --> It all started from ignoring PAR and seeing 720/405 as 16:9, which it isn't. As it removes less lines of resolution, it's just a nice starting point.]

Now, ready for this ? I hit on the following computations somewhat by accident, but I have reasonable hope that they are correct:
my goal is to get to
720(1.2)x480 -- wide DV footage
720(.9)x405 -- my frame, which I need to adapt to wide footage

need to resize ...
720(0.9)x405 current frame (16:9)
x(1.2)x480 compute how it will be like in wide footage

x=640

Using MainConcept I did this:
Crop to 38 pixels from both top and bottop (thus giving me 720x404 frame. 404 not 405 as encoders like even numbers).
Resized footage after cropping to 640x480.

[jagabo]

Starting with 720x480 4:3, cut 60 lines off the top, 60 lines off the bottom, resize what's left (720x360) to 720x480. Encode as 16:9.

[dennismv]

thank you. Yes that'll work best technically, but I didn't want to lose 120 lines of resolution. With the other method I lose just 75. I'll give it a try but ultimately if I have to cut off that much, I'd rather keep it at 4:3.

[Cornucopia]

You're forgetting--
1. Pixel Aspect Ratios
2. 704 vs. 720 dead L, R edge scan areas.

So:

704 / 480 *.909090909 (NTSC PAR) = 4:3 DAR ( 1.333333 )

704 / 360 *.909090909 (NTSC PAR) = 16:9 DAR ( 1.78 )

***Those extra 8 pixels on each side of a DV frame (L, R) don't count toward DAR (They're supposed to be black)***


What you've got:

720 / 405 *.909090909 (NTSC PAR) = 1.6161616 ??

You CAN'T just ignore PAR. It won't just "go away".

Plus, "auto resizing" is all about ANAMORPHIC encoding (and authoring). You encode your cropped (720x360) footage stretched(resized) to 720x480 and then authored AS anamorphic 16:9 (you have to tell the authoring app to do that).

Notice:
4:3 DV footage is 720x480
16:9 (anamorphic) DV footage is ALSO 720x480.
It's equivalent resolution could be called 853x480 or 720x360 (depending upon whether you consider it stretching or shrinking).

jagabo had good advice that 1st post...

Scott

[dennismv]

yes that's all correct. (though I think PAR is .9 as per specs. I tried to verify it but couldn't find an authoritative source)
I didn't want to throw out 120 lines of resolution.

So I got inventive. I have included PAR, but I did a few manipulations along the way that are detailed in the first post. Best way to show it will be via a picture. Though I don't want to draw one right now as it's still easier (for me) to explain in words:

720 / 405 is not my final resolution, but it is a temporary output file I've used. This file only throws out 75 lines of resolution compared to 120. I tried throwing out 120 and it was too much -- important elements were getting cut off.

So what I did then was to encode the file to MPEG2, resizing it in the process detailed in first post. Result is not a true 16:9, but a 16:10 (or 14.4:9 to be more precise)
720*.9/405 = 14.4/9

This means the file plays with two black bars on the sides. I'm content with that as it's a compromise between getting close to 16:9 and not throwing away a lot of footage in the process.

[Cornucopia]

Your history-revisionist efforts notwithstanding, you did say in the OP:

What I have: 4:3 DV footage
What I want to do: get a 16:9 footage

What you did DOESN'T give you 16:9.

You stated yourself that going fully to 16:9 cuts off too much stuff, which I would agree with. In fact, I'd go so far as to say you should only cut off 0 pixels ( aka, leave it as it is ).

If you're going to change the majority of your post, why not change the subject(thread) title to read: "for 4:3 DV to 14.4:9 DV"?

Scott

[NY2LA]

You're forgetting--
1. Pixel Aspect Ratios
2. 704 vs. 720 dead L, R edge scan areas.

So:

704 / 480 *.909090909 (NTSC PAR) = 4:3 DAR ( 1.333333 )

Scott

***Those extra 8 pixels on each side of a DV frame (L, R) don't count toward DAR (They're supposed to be black)***

Scott

This is exactly what I've been thinking, when cropping and scaling DV to a custom size of 640 x 480 during export to H.264. (Cropping 8 pixels off the left and right of the source, and horizontally scaling to 640 by entering a custom size of 640 x 480.) But I am dealing with two mysteries:

1.) Why, when I look at DV with MediaInfo, I see a Pixel Aspect Ratio of 0.889, not .909? Even inside a Quicktime wrapper, MediaInfo will say PixelAspectRatio_Original: 0.889.

2.) Why does Charles Poynton, in his Digital Video and HDTV Algorithms and Interfaces, 480i component video chapter, say: In 4:3 systems, aspect ratio is defined to be 4:3 with respect to a clean aperture pixel array, 708 samples wide at sampling rate of 13.5 MHz, and 480 lines high." (Page 505, and illustrated in Figure 41.2 on page 507.) Why 708 samples wide rather than 704?

[Cornucopia]

@NY2LA,
Normally, its considered quite bad form to resurrect a 5 year old thread in such a way, but it's Christmas so I'll ignore that for now....

In answer to your 2 mysteries:

1. Some follow the Rec601 guidelines for screen digitization/composition/AR and some follow the DTV/post-dvd-digital guidelines, when doing their calculations. I usually follow REC601, so I use 704 as the "core" with 720 used for overscan. Use of a 4:3 dar with this expects a 0.9090909 par.
Those that follow the dtv/post-dvd-digital guidelines use the full 720 width in their calculations, so they arrive at a par of 0.88888888.
There are benefits and drawbacks to both approaches...

Obviously, MediaInfo and QT seem to be referring to the 2nd approach (although QT has a lot more quirks...)

2. Don't know - I don't have a copy of that text. Could be a typo. Could be that poynton was describing the whole difficulty with trying to arrive at par in the 1st place. By many accounts, for an actual TRUE 4:3 digital dar to be digitized from an analog 4:3 signal, one would have to use 711.??? as the horizontal for NTSC and 702.9??? for PAL. For expediency's sake and to make it easier for encoders to do their business (with mod16), both numbers were rounded to 704. Poynton is certainly aware of the circuitous historical route that our current methods of creating digital media have undergone.

Scott

[NY2LA]

@NY2LA,
Normally, its considered quite bad form to resurrect a 5 year old thread in such a way, but it's Christmas so I'll ignore that for now....

Scott

Sorry, I didn't realize this. Thanks for telling me. I honestly thought that it was a good idea, because so many questions and answers get repeated; so many new threads get started on subjects that have already been covered.

I'll reply later to your other replies, though.

[NY2LA]

Correction: I saw the PAR with DVAnalyzer, not MediaInfo. The author of DVAnalyzer once wrote to me: "DVAnalyzer uses Mediainfo to parse the container . . .", so that's what made me think of MediaInfo.

[NY2LA]

@Cornucopia

Re: your answers to my two mysteries:

1.) Very helpful. Thanks.

2.) Poynton says the same thing in his earlier book, A Technical Introduction to Digital Video, in Chapter 12, "525/59.94 scanning and sync", on page 213: "Aspect ratio is defined to be 4:3 with respect to a clean aperture pixel array, 708 samples wide at sampling rate of 13.5 MHz, and 480 lines high." Also, in Chapter 13, "525/59.94 component video", page 217, Figure 13.2, "525/59.94 component digital 4:2:2 luma waveform", shows "708 clean". The context is about blanking transition samples, which he says "should not intrude on the clean aperture". I don't see details about "the circuitous historical route" to arriving at PAR in either book. (I wish he would write about it.)

[jagabo]

The DVD spec refers to the MPEG 2 spec regarding aspect ratios. The MPEG 2 spec is very clear: the full 720x480 frame contains the 4:3 or 16:9 picture (unless overridden by a display sequence extension -- which I've never seen except to define a 4:3 frame for pan-and-scan on 16:9 DVDs). Not the inner 704x480 (ie the ITU/BT spec). That makes the PAR on 720x480 4:3 NTSC DVD 8:9, not 10:11 (ITU/BT 601).

[NY2LA]

The DVD spec refers to the MPEG 2 spec regarding aspect ratios. The MPEG 2 spec is very clear: the full 720x480 frame contains the 4:3 or 16:9 picture (unless overridden by a display sequence extension -- which I've never seen except to define a 4:3 frame for pan-and-scan on 16:9 DVDs). Not the inner 704x480 (ie the ITU/BT spec). That makes the PAR on 720x480 4:3 NTSC DVD 8:9, not 10:11 (ITU/BT 601).

Very interesting. I wonder if this means many video editors may be unaware of the 10:11 (ITU/BT 601) spec. Doesn't this also imply a short cut export feature (i.e., Apple's Share to Apple Devices . . .) is going to have to follow one or the other spec, and so it may do it incorrectly with some video?

[jagabo]

I don't know who's aware of what. But when examining actual DVD VOB files (looking for objects of known aspect ratio and measuring them on-screen) some seem to follow the MPEG spec (h.262), some the ITU/BT spec. In any case the difference is only about 2 percent. So it's not really noticeable. And it was much less than the average AR error of CRT displays (if you remember those!).

[NY2LA]

I don't know who's aware of what. But when examining actual DVD VOB files (looking for objects of known aspect ratio and measuring them on-screen) some seem to follow the MPEG spec (h.262), some the ITU/BT spec. In any case the difference is only about 2 percent. So it's not really noticeable. And it was much less than the average AR error of CRT displays (if you remember those!).

The 2 percent difference may not be noticeable, except when it consists of thin black bars, particularly with an analog source, I think.

[jagabo]

The 2 percent difference may not be noticeable, except when it consists of thin black bars

You mean moire artifacts from scaling?

[NY2LA]

The 2 percent difference may not be noticeable, except when it consists of thin black bars

You mean moire artifacts from scaling?

No, I don't mean moire artifacts from scaling. I mean, they are visible, rather than cropped off. That's all.

[jagabo]

The 2 percent difference may not be noticeable, except when it consists of thin black bars

You mean moire artifacts from scaling?

No, I don't mean moire artifacts from scaling. I mean, they are visible, rather than cropped off. That's all.

You mean the "extra" 8 pixels at the left and right edges? Actually, most TVs won't show them because of overscan. On TVs that don't overscan those columns should be visible because they are part of the 4:3 or 16:9 picture. Technically, a 704x480 D1 video should be stretched to 720x480 for DVD. A 720x480 D1 video should have 8 pixels at the left and right edges cropped and the remaining 704x480 stretched to 720x480. But I think most DVD productions that start with analog video don't bother. At most they will overlay black bars to remove any noise that's out there.

[NY2LA]

You mean the "extra" 8 pixels at the left and right edges?

Yes, I mean the "extra" 8 pixels at the left and right edges. The thing is, I am scaling during export to a custom size of 640 x 480. I think they should only be visible if I was scaling to 654 x 480.

[jagabo]

You mean the "extra" 8 pixels at the left and right edges?

Yes, I mean the "extra" 8 pixels at the left and right edges. The thing is, I am scaling during export to a custom size of 640 x 480. I think they should only be visible if I was scaling to 654 x 480.

No, on a DVD the 4:3 image is contained in the full 720x480 frame.

[NY2LA]

No, on a DVD the 4:3 image is contained in the full 720x480 frame.

I am working with DV, not DVD. (See my earliest post above.)

That's good to know about DVD's, though, for future reference.

[intracube]

The DVD spec refers to the MPEG 2 spec regarding aspect ratios.

Do we know that for sure?

Which DVD spec refers to the MPEG 2 spec?

That was one of the roadblocks in the original discussion over 702 vs 720. DVD standard = $$$ + non disclosure.
http://www.dvdfllc.co.jp/license/l_howto.html

[jagabo]

The DVD spec refers to the MPEG 2 spec regarding aspect ratios.

Do we know that for sure?

In one of the threads here a few years ago someone with a hard copy of the DVD standard checked. It referred to the MPEG 2 spec for aspect ratio information.