[opendtv] Re: Finally anamorphically compressed 480i

  • From: Ron Economos <w6rz@xxxxxxxxxxx>
  • To: opendtv@xxxxxxxxxxxxx
  • Date: Tue, 01 Feb 2011 04:51:43 -0800

You're essentially correct that ATSC MPEG-2 encoders don't
use field based encoding. Field encoding (at least with MPEG-2)
is usually only done on low-delay encoders.

However, ATSC MPEG-2 encoders (even old ones like
the Harris Flexicoder) select field or frame based
encoding on a macroblock by macroblock basis (also
known as MBAFF). Here's a B-frame from the local CW
affiliate (looks to be a fairly new Harris NetVx encoder).

The luma signal showing that it's an interlaced frame.

B-frame luma


Here's the macroblock type for the same frame. Blocks with a
line though them are using the field DCT. Forward predicted blocks
are blue, backward yellow and bi-directional are pink. A dot is
used to show when the quant level changes.

B-frame mtype

Here are the motion vectors. The blocks with an upper and
lower partition are field predicted. For bi-directional blocks,
there are four vectors per macroblock.

B-frame motion vectors

Ron

On 1/31/2011 7:12 AM, John Shutt wrote:
Do you really want me to buy that?  Once in an MPEG-2 stream, the content is
frame based.

And I'm talking about practical day-to-day broadcast operations using real world
ATSC encoders, not theoretical encoding tools in a specification, or how anyone believes it "should" be done.  I'm dealing in real world hardware.

In the same way that ATSC theoretically allows separate audio voiceovers to be mixed into the main audio channel, or
for receivers to act upon a Directed Channel Change, in the real world, with real equipment, it cannot be done.

All of the MPEG-2 encoded material we deal with, be it DVB-S transport streams from PBS, video clips ingested into an Omneon video server from an SD-SDI video source, or our ATSC program stream created by Tandberg encoders, is frame based, and most of it uses a 15 frame GOP.  No selective field encoding, and the only exception to the GOP sequence is scene cut detection that will force an I frame early.

John

----- Original Message ----- From: "Craig Birkmaier" <craig@xxxxxxxxx>
To: <opendtv@xxxxxxxxxxxxx>
Sent: Saturday, January 29, 2011 11:25 AM
Subject: [opendtv] Re: Finally anamorphically compressed 480i

Not exactly...

Look at the H.262 specs and you will find all kinds of tools for encoding
native interlaced source.

For example:

4.1.2 Coding interlaced video
Each frame of interlaced video consists of two fields which are separated
by one field-period. The specification allows either the frame to be
encoded as picture or the two fields to be encoded as two pictures.  Frame
encoding or field encoding can be adaptively selected on a frame-by-frame
basis.  Frame encoding is typically preferred when the video scene
contains significant detail with limited motion.  Field encoding, in which
the second field can be predicted from the first, works better when there
is fast movement.


Other related posts:

  1. Home
  2. opendtv
  3. Archive
  4. 02-2011