[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
vm_lowdown from Scitex
The following is submitted by David Workman in participation with the
'vm_lowdown' program that helps support this group. For more
'http://www.alegria.com/telecine/info.html/#vm_lowdown', or email
rob at alegria.com.
--- Forwarded mail from David Workman <workmand at abekas.com>
Abekas is preparing to ship version 3 software for Dveous, which has
one feature that the telecine group might be interested in: true frame
rate image manipulation. Dveous has four field framestores on it's
inputs (unlike most DVE's, which only have a three field framestore),
which is of particularly use for manipulating video coming from film
With a three field framestore, you are always filling one field of
video while interpolating between the previous two. The previous two
fields, however, are always swapping between coming from a single
frame, and coming from field 2 of one frame and field 1 of the next.
The result is that every other frame is of a lower quality when you
put frame rate material through the DVE.
Some of you (mostly higher end users) may be familiar with "doubling".
This is where you run the film through the telecine at half speed, so
you get two frames of video for every frame of film. When you run
this through the DVE (in a typical 3 field store system) every other
frame is interpolated, as described above, and then you "throw out"
every other frame of the resulting video to get back to normal speed.
A four field framestore doesn't have this problem, however, because
it's always filling two fields while interpolating between two fields
of a single frame. The phase of this can be changed, so that it can
span the frame boundary if the video has somehow been offset by a
field in some other processing. In 3:2 mode, there's a 5 field
sequence and you can start the processing on any one of the 5 fields
(in reality, it's a 20 field sequence that repeats four times, but
that's hidden from the operator!)
The processing modes, described above, are different from the motion
control. In past DVE's (K-Scope, A57, etc) you had a FRAME and 3:2
selection, but it only adjust the motion of the transform engine
(updating the position of the picture once every frame or on a 3:2
sequence). You need to do BOTH (video processing and transform update)
to properly handle frame rate material.
As part of this processing, Dveous also upsamples the video and
de-interlaces it. The A57 did a "horizontal" upsampling, filling in
every other pixel, making a 27MHz bit stream. Dveous does a
"vertical" upsample, creating a full frame of video out of every
field. In both cases, the net result is better picture quality, since
you're feeding the transform engine twice as much video information.
In Dveous, however, you get the best of both worlds with 4 field
processing and superior picture quality.
--- End of forwarded message from David Workman <workmand at abekas.com>
Rob Lingelbach | 2660 Hollyridge Dr., Los Angeles, CA 90068
rob at alegria.com | "I care not much for a man's religion whose dog or
rob at sun.alegria.com | cat are not the better for it." --Abraham Lincoln
rob at praia.alegria.com KB6CUN http://www.alegria.com