[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re; Layering vrs Windowing
- To: "Rob Lingelbach" <rob at sun.alegria.com>
- Subject: Re; Layering vrs Windowing
- From: "Seamus O'Kane" <spok at compuserve.com>
- Date: Sat, 10 May 1997 14:17:21 -0400
- Old-Return-Path: <rob>
- Organization: Altruistic Intentions, Hollywood, CA
- Phone-number: +1 213 464 6266
- Reply-To: Rob Lingelbach <rob at alegria.com>
- Resent-Date: Sat, 10 May 1997 12:07:38 -0700
- Resent-From: rob at sun.alegria.com (Rob Lingelbach)
- Resent-Message-Id: <m0wQHUY-000AKaC at sun.alegria.com>
- Resent-Sender: telecine-request at sun.alegria.com
- Resent-To: telecine internet group <telecine at alegria.com>
To avoid the "marketing" issue I am offering this posting via
Rob and the views are all mine as a user.
Bob Festa posted;
>colorist friend Seamus went into great detail at the show, however my mind
>was boggled with other items of interest.
>Perhaps someone who has spent time with both systems can explain the
>differences in attacking an isolation challenge.
As Bob has mentioned my name it would seem churlish not to
offer some reply on this subject and also make some amends
for not getting all of the information over at the show. I cannot
respond to the second part of his question though so I will
just offer a brief description of how "layering" and "windows"
relate in the DCP/ESR .
DCP/ESR has twelve channels of colour corrrection and a set of
overall master primary controls. There are two secondry
colour correctors, the six channels of each are conventionally
identified as R,G,B.Y,C,M. Whilst both can be used in this fixed
manner the variable isolation of each, especially DCP 2, allows
for greater flexibility. DCP 1 is a "simple" six vector device based
on hue isolation, whilst DCP 2 is a "complex " corrector which
uses hue, saturation and luminance to isolate objects or areas
within the picture.
In addition to this full isolation, DCP 2 has a variable priority to the
channels which introduces the concept of layering, ie it is possible
to use an uncorrected isolation from a "higher" layer to assist in the
isolation and correction of a "lower " layer, or more typically, an
overall colour effect such as a tint, may be introduced on the lowest
layer and original colours brought back on the layers above, say a
red dress on one and green trees on another etc,etc. Once
they have been isolated and associated with a layer their
original colours can be manipulated. All six of the layers on DCP 2
have hue, saturation and luminance output controls and in addition
three of them also have a full set of primary controls.
In the real world the successful isolation of these colours depends
on a host of changing conditions. The hue, saturation and
luminance values of the isolation can be softened,
independently for each, to enable these natural variations to be
less obvious and the output colours to be more natural. The three
layers with primary controls can also have the output of those
faded back into the lower layers and the background.
So to use the layering principal it is a question of deciding what
effects are needed once the picture is brought to an overall
balance. It may be that only one layer is needed to change one
object or several can be used to change many. The principal
is the same in that a layer is selected and the isolation is used
to group pixels with the same value of hue, saturation and
luminance, and then affect those values at the output stage.
It is also possible to use a "region" or "window" in tandem with
the isolation, as each of the six layers on DCP 2 can enable it's
own softedged shape. These always start from a rectangle and
can be dyamically altered and positioned at will. This can make
it possible to use the contained primary controls of the top three
layers as light sources within the overall balanced picture and, if
isolation is used within a soft region, it is possible to make very
photographic corrections and effects. As well as creating an image
that would previously have been composed from multi passes,
working with regions and layers is quick and therefore allows
for several solutions to be offered for Client approval.
The problem with "layers & windows" is that the real world can
so easily fool both. It is impossible to rely on just one method
and the more interactive they are the better. It is quite common
to use my three isolation variables to define an object and to
quickly throw a region around it as it moves but I do not wish
to get involved in complex drawing or keyframe plotting. We
have Flame / Henry rooms that are far better suited and can
cope with the changing requests of a room full of Clients. I
don't think it pleases anyone to watch you struggle for ages
tracking numerous windows and then the Art Director asks
for a 30 percent blowup and some rotation.
By using the three colour parameters to isolate I am often
able to proceed without a region and, if I soften the isolation
enough, then again, that helps with the intergration of the
result. If I do need to use a layers region then its variable
softness will reduce the need for intense tracking by producing
a vague area of containment. In the same manner I find that
changing the colour of the isolated and, (or), contained region
is much more natural when done with a set of primary
controls rather than conventional secondry hue, saturation
and luminance. Introducing the change and matching gamma
to the surrouding picture seems to be the key here especially
on extreme changes.
The next step must be to increase the isolation ability beyond
colour parameters as there is little point in attempting more
complex corrections just by adding in regions, windows or
layers. RSQ handles its multi layers very well in they can be
ganged and are very easily assignable but it gets very difficult
to envisage an interface and control system that would go
beyond these 32 layers. In the same manner increasing the
amount of regions, on a layer or overall, is not really helpful
unless the contents can be used independently.
In a similar fashion it must soon be that an object defined
by full colour parameters, and any other means, is able
to be identified and tracked as the scene progresses
without the need for repeated keyframing. I would
have thought that the layering approach would be more
suited to this type of advanced image tracking as it
seperates elements in a more logical way.
So, although I cannot offer recent experience of any other
system, I hope that this may have thrown some more
light on the "layering" approach of the DCP. The whole
system has become quite large now and the difficulty
at a show is to try and explain everything in quite a short
time. Sorry Bob if I failed, but thanks for such an eloquent
VTR Ltd . London.
thanks to Ken Rockwell, Dwaine Maggart, and Joe Wolcott
for support of the TIG in 1997
mailinglist digest available......posting guidelines on the webpage