[opendtv] Re: News: See the Big Picture? Don't Forget to Examine the Fine Print
- From: Craig Birkmaier <craig@xxxxxxxxx>
- To: opendtv@xxxxxxxxxxxxx
- Date: Tue, 4 May 2004 16:42:43 -0400
Please define "screen aspect ratio." This is not a term that is used
in the MPEG-2 context.
Are you trying to say that the sample aspect ratio is NOT 1:1 (i.e. square)?
Regards
Craig
I would also add, that it is possible to signal almost anything in
the private data area of the MPEG-2 transport standard.
At 12:50 PM -0700 5/4/04, John Willkie wrote:
>Really? it doesn't limit? Then, pray tell (note the tone, if you call that
>rude, well, that's a clue) please indicate how, using the existing
>mechanisms, I would indicate a screen aspect ratio of 1:2.88 with a display
>aspect ratio of 1:0.72. Note: I made it "somewhat" easy, as 0.72 can be
>easily factored into 1:2.88.
>
>But, you say that MPEG-2 is flexible. So, tell me how I would signal such a
>combination using the MPEG-2 infrastructure.
>
>And, if you can't, rather than just provide material that proves MY POINT,
>would you for once concede?
>
>John Willkie
>
>-----Original Message-----
>From: opendtv-bounce@xxxxxxxxxxxxx
>[mailto:opendtv-bounce@xxxxxxxxxxxxx]On Behalf Of Craig Birkmaier
>Sent: Tuesday, May 04, 2004 11:57 AM
>To: opendtv@xxxxxxxxxxxxx
>Subject: [opendtv] Re: News: See the Big Picture? Don't Forget to
>Examine the Fine Print
>
>
>At 10:00 AM -0700 5/4/04, John Willkie wrote:
>>You are VERY incorrect.
>
>Really?
>
>I will disregard this statement and its rudeness,=20
>as i do much of what you spout in this forum.
>
>=2EI've worked in the MPEG process for years and I=20
>know exactly of what I speak. If you were to have=20
>included the actual text from the MPEG-2 standard=20
>everyone would see just how wrong you are.
>
>
>
>>MPEG-2 provides several mechanisms that signal video size ans aspect ratio
>>(section 1: systems), including a flag that signals a video frame as being
>>1:1.33; 1:1.78 and 1:2.21, and the size relative size of the picture
>>elements. Look up the extended segment.
>
>Correct. From section 6.3.3 of the standard (I do=20
>not have the latest version, but I do not think=20
>this has been amended). I am only quoting the=20
>relevant parts of this section:
>
>6.3.3 Sequence header
>
>horizontal_size_value -- This word forms the 12=20
>least significant bits of horizontal_size.
>
>vertical_size_value -- This word forms the 12=20
>least significant bits of vertical_size.
>
>horizontal_size -- The horizontal_size is a=20
>14-bit unsigned integer, the 12 least significant=20
>bits are defined in horizontal_size_value, the 2=20
>most significant bits are defined in=20
>horizontal_size_extension. The horizontal_size is=20
>the width of the displayable part of the=20
>luminance component of pictures in samples. The=20
>width of the encoded luminance component of=20
>pictures in macroblocks, mb_width, is=20
>(horizontal_size + 15)/16. The displayable part=20
>is left-aligned in the encoded pictures.
>In order to avoid start code emulation=20
>horizontal_size_value shall not be zero. This=20
>precludes values of horizontal_size that are=20
>multiples of 4096.
>
>vertical_size -- The vertical_size is a 14-bit=20
>unsigned integer, the 12 least significant bits=20
>are defined in vertical_size_value, the 2 most=20
>significant bits are defined in=20
>vertical_size_extension. The vertical_size is the=20
>height of the displayable part of the luminance=20
>component of the frame in lines.
>In the case that progressive_sequence is '1' the=20
>height of the encoded luminance component of=20
>frames in macroblocks, mb_height, is=20
>(vertical_size + 15)/16.
>In the case that progressive_sequence is '0' the=20
>height of the encoded luminance component of=20
>frame pictures in macroblocks, mb_height, is=20
>2*((vertical_size + 31)/32). The height of the=20
>encoded luminance component of field pictures in=20
>macroblocks, mb_height, is ((vertical_size +=20
>31)/32).
>The displayable part is top-aligned in the encoded pictures.
>In order to avoid start code emulation=20
>vertical_size_value shall not be zero. This=20
>precludes values of vertical_size that are=20
>multiples of 4096.
>
>aspect_ratio_information -- This is a four-bit=20
>integer defined in the Table 6-3.
>
>aspect_ratio_information either specifies that=20
>the "sample aspect ratio" (SAR) of the=20
>reconstructed frame is 1,0 (square samples) or=20
>alternatively it gives the "display aspect ratio"=20
>(DAR).
>
>* If sequence_display_extension() is not=20
>present then it is intended that the entire=20
>reconstructed frame is intended to be mapped to=20
>the entire active region of the display. The=20
>sample aspect ratio may be calculated as follows:
>
> SAR=3DDAR x horizontal_size/vertical_size
>
>NOTE - In this case horizontal_size and=20
>vertical_size are constrained by the SAR of the=20
>source and the DAR selected.
>
>* If sequence_display_extension() is=20
>present then the sample aspect ratio may be=20
>calculated as follows:
> SAR=3DDAR x
>display_horizontal_size/display_vertical_size
>
>
>
>Table 6-3 - aspect_ratio_information
>aspect_ratio_information Sample Aspect Ratio DAR
> 0000 forbidden forbidden
> 0001 1,0 (Square Sample) -
> 0010 - 3=F74
> 0011 - 9=F716
> 0100 - 1=F72,21
> 0101 - reserved
> =BA =BA
> 1111 - reserved
>
>---------
>
>Now, what does all of this mean.
>
>=46irst, the standard accommodates two different=20
>meanings of aspect ratio, which together can=20
>influence the actual display aspect ratio. First,=20
>it considers the sample aspect ratio (SAR);=20
>second it considers the display aspect ratio DAR.
>
>John is correct that the standard does provide=20
>values for the common 4:3 and 16:9 aspect ratios=20
>as well as 2.21. But this does not limit the=20
>ability to designate other SARs or DARs.
>
>=46or example, How does MPEG deal with ITU-R BT.601 samples?
>
>The size of the raster is stated in=20
>horizontal_size and vertical_size. Let's say it=20
>is the full 720 x 480/576. But are these samples=20
>representing a 4:3 or 16:9 raster. You must look=20
>at aspect_ratio_information to figure this out.=20
>You can also use any fractional line length in=20
>Horizontal size - the display will then stretch=20
>the sample to fill the display aspect ratio.
>
>=46ortunately "601" is the only video standard in=20
>use today that uses variable sample aspect ratio=20
>9actually the compressed HD camcorders do this=20
>internally, but they always produce square=20
>samples at the output.
>
>Now how would I designate something other than 4:3 or 16:9 or 2.21:1?
>
>Simple. I would use horizontal and vertical size=20
>to designate the raster size, and 0001 in aspect_=20
>ratio_information to designate square samples.=20
>Voila... I can represent ANY aspect ratio. It=20
>gets a bit more tricky if I do not use square=20
>samples; I would then need to use a reserved=20
>extension to indicate the DAR and calculate the=20
>SAR using the formulas.
>
>John must have been confused by the fact that=20
>MPEG took the time to include values for the=20
>standards that exist. But this does not mean that=20
>you cannot use the standard as I have described.
>
>
>>
>>ATSC supports only specific video formats, and extended segment support is
>>specified.
>>
>>DVB-SI provides mechanisms -- in addition to MPEG-2 section 1 -- to signal
>>video frame aspect ratio.
>>
>>Wanna try again?
>
>Why don't you give up before you make an ass out of yourself again.
>
>Regards
>Craig
>
>
>----------------------------------------------------------------------
>You can UNSUBSCRIBE from the OpenDTV list in two ways:
>
>- Using the UNSUBSCRIBE command in your user configuration settings at
>FreeLists.org
>
>- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
>unsubscribe in the subject line.
>
>
>
>
>----------------------------------------------------------------------
>You can UNSUBSCRIBE from the OpenDTV list in two ways:
>
>- Using the UNSUBSCRIBE command in your user configuration settings
>at FreeLists.org
>
>- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the
>word unsubscribe in the subject line.
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:
- Using the UNSUBSCRIBE command in your user configuration settings at
FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
unsubscribe in the subject line.
Other related posts:
