🚀 go-pugleaf

In article <1992Jul19.153254.22747@becker.GTS.ORG> bdb@becker.GTS.ORG (Bruce Becker) writes:
>In article <4574@crl.LABS.TEK.COM> rjn@snowbird.LABS.TEK.COM (Jim Nusbaum) writes:
>|Well CIF is supposed to be half the size of the full image (dropping every
>|other sample in the line and dropping alternating lines).  That would lead to
>|320 x 240 starting from a 640 x 480 image.  The 4:1:1 subsampling leads to
>|the 1/4 size U and V.  If you start with a different image size you would
>|get different results, but a 704 x 480 image seems pretty non-standard.
>
>
>	704 == 640 + 64
>
>
>	the extra 64 is the additional 10%
>	needed for overscan
>
>
>-- 
>  ,u,	 Bruce Becker			Toronto, Ontario
>a /i/	 Internet: bdb@becker.gts.org	Uucp: ...!web!becker!bdb
> `\o\-e	 \*/*\*/*\*/*\*/*\*/*\*/*\*/*\*/*\*/*\*/*\*/*\*/*\*/*\*/
> _< /_	 "Awaken the Geek Within" - Anthony Robot


Well, the number 704 was actually derived from the follwing reasoning:
	CCIR601 format has 720 pixels per line
	--> Decimate it down to 360 pixels
	--> To divide the frame into macroblocks ( 16 x 16 pixels),
		there is 8 pixels as the remainder (360 /16 = 22 remainder 8)
	--> Now we just discard the 8 pixels and use the 352 pixels for
		the MPEG encoding. In fact we take the center 352 pixels
		discarding 4 pixels from both sides.
	--> The above procedure effectively takes the 704 center pixels
		from the original frame (Of course you may need the 
		boundary 16 pixels to do the decimation).

This is illustrated in the MPEG document.

Dong-Lok Kim
-- 
ICSL (Image Computing System Lab)
Dept. of Electrical Eng., FT-10
University of Washington
Seattle, WA 98195