On Mon, 26 Apr 2004 15:29:02 GMT, "David Ruether"

<rpn1@no-junk.cornell.edu> wrote:

>"Five" <Niko@fiveminutesof_blank.com> wrote in message

>news:Qj9jc.41636$aQ6.2382411@attbi_s51...

>> The biggest problem I see is that it is HDV. That is

>> taking a larger image, and compressing it more, and

>> also doing so in a lossy format to the same size of

>> capture media (DV tape).

>Actually, the two compression types are quite different,

>with different results.

Yup.

>As I understand it, Mini-DV is frame-by-frame MPEG1,

Not really. MPEG-1/2 I-Frames are in the same basic family as DV

frames, but they're not the same. DV is closer still to Motion JPEG,

but [naturally] a more rigid standard (MJPEG is kind of an ad-hoc

thing), with optimizations for interlaced video, like the MPEG folks

introduced in MPEG-2. All are the same basic family, using a

reversible discrete cosine transform to change pixel blocks, or

macroblocks (sizes depend on the spec, but usually 8x8 or 16x16

pixels, though modern formats like H.264 have a variable block size)

of spatial information into pixel blocks of frequency information,

which is then quantized (low-pass filtered in spatial terms) and then

compressed losslessly with something like Huffmann encoding. It's the

quantization pass that results in the loss, for this kind of

compression. MPEG can use different quantizing factors for each

macroblock, not done in MJPEG and I don't believe in DV, resulting in

a 10-20% density improvement, even in I-Frame-only video.

> with a data rate that provides one hour of DV-SD compressed

> 5:1 to fit on Mini-DV tapes (SP-mode).

Yup. That's about a 12GB tape. And naturally, as DV is compressed

frame by frame, it's very thing to edit, as each frame is like every

other, and independent of every other.

>MPEG2 is not frame-by-frame, so it

>is easier to compress a higher-resolution image to less than

>the Mini-DV SP-mode data rate with picture quality that

>is in some ways superior in terms of visible compression

>artifacting (though this compression type is harder to edit

>easily).

Yup. MPEG-2 is actually compressed in units called GOPs (Group of

Pictures). Each GOP starts with an I-Frame (Independent Frame), that's

more or less like a DV or MJPEG frame. GOPs also include two other

kinds of frames: P-Frames and B-Frames.

A P-Frame (predictive frame) can follow an I-Frame, and this is now

changing from the spatial redundancy of JPEG/DV to temporal

redundancy. It's usually the case that one frame is very much like the

next. So when you're encoding that next frame, various algorithms

figure out motion vectors: where each macroblock in the I-Frame can be

found in the current frame, to a 1/2 pixel accuracy. The P-Frame,

thus, only needs to store these vectors, and the error terms: the real

difference between the I-Frame and the current frame. The P-Frame is

usually dramatically smaller than the I-Frame, but of course, it

doesn't existing independently, you have to have the I-Frame.

Next is the B-Frame (bidirectional predictive frame) can be predicted

from I-Frames or P-Frames, both forward and backward. Otherwise,

similar to the P-Frame.

There are lots of resources on MPEG-2 if you want to learn more, or

want a better explanation.

>Also, as I understand it, HDV is either 1080i

>(with all the motion-advantages of interlacing), or 730p

>(with a PS frame rate of 30, too slow to avoid motion

>flickering easily).

720/30p, actually. That's 1280x720 at 30fps, progressive. As mentioned

before, the ATSC/HDTV standard is actually 720/60p. This is going to

be jerky, like film. Whether that's a problem or not, I guess, depends

on whether you wanted film or video. If they're really targeting film

replacmenet with these cameras, this can't really be too much of an

issue, but they should have a 24p mode, too.

Real visual flicker won't be an issue; no one would actually view this

at a hardware 30fps. Assuming ATSC systems are happy with 720/30p (I

think it's on the list -- a total of 18 formats are supported within

ATSC), they would upconver to either 720/60p or 1080/60i.

>> We should at least get a larger tape even if that

>> requires us to use large DV tapes. IF size is of

>> concern, then using a new formulated 8mm tape would

>> offer longer potential run times due to the ability to

>> put more tape in the shell. You are creating a new

>> format, so why not a new tape size and formulation.

>The HDV standard, as I understand, was set with

>the original Mini-DV specs - but there has been a

>considerable delay in its implementation.

I rather doubt it. For one, I don't think the whole "Grand Alliance"

had settled on everything that was going into HDTV by the time the DV

specs were set in stone. And then there's the fact that the two JVC

camcorders came out BEFORE the industry got together and dubbed this

HDV. And then there's this from www.hdv-info.org:

Tokyo, Japan, September 30, 2003 - Canon Inc., Sharp

Corporation, Sony Corporation, and Victor Company of Japan,

Limited (JVC) today announced that the specifications which

realize the recording and playback of high-definition video on

a DV cassette tape have been established as the "HDV" format.

The four companies proposed the basic specifications for the

format in July 2003 and have received supportive opinions from

many companies. The specifications for the HDV format will

become available from the beginning of October 2003. The four

companies will also propose the format as an international

standard format.

Dave Haynie | Chief Toady, Frog Pond Media Consulting

dhaynie@jersey.net| Take Back Freedom! Bush no more in 2004!

"Deathbed Vigil" now on DVD! See http://www.frogpondmedia.com