The GEOMETRY Workflow - 3D Postproduction: Stereoscopic Workflows and Techniques - page 125

4.7 Real-Life 3D Post: GEOMETRY
by Juan I. Cabrera
4.7.1 Overview
Factors such as the type of 3D rig used for shooting, the experience of the
crew and how much time there is to adjust the alignment for each shot will
dramatically affect the amount of correction you have to perform. As a rule of
thumb, the more complex the rig the more attention and expertise it requires.
If you have any input in the decision of choosing the 3D gear to be used,
take this into account. Sometimes it is worthwhile to go to something more
basic (like an all-in-one 3D camera) if there isn't going to be enough time to
shoot it correctly with the more complex rig.
In 99.9% of the cases when you encounter geometry issues, there won't be just
a single issue, but a combination of them (Position, Scale, Rotation, Keystoning,
for example). There are multiple tools in the market that feature automatic and
semi-automatic geometry correction. Some of them are truly amazing, accurate
and fast. But you will always i nd a shot (or a number of them) where the
“magic button” doesn't work. For those cases, and to be able to tweak what
the tool does, you need to know what is happening with the images and
how to i x it. For that, I offer you this basic i ve-step process that should get
you “in the ballpark” in most cases. Learn how to identify them and how to
correct them. Best of all, we will view our 3D image pair in 2D for a change!
To illustrate some of the key geometry topics in this section, we use a frame
from an S3D show provided by Ned Weisman. Here is what the frames look
like prior to viewing them in difference mode (Figures 4.18 and 4.19) .
4.7.2 Position
Let's take the 3D glasses off and forget about convergence for a second. To
help you with this you can use any kind of visualization mode that allows
you to clearly see both images simultaneously in 2D as Transparencies, B&W
Anaglyph, Difference. Line up L and R images (both in X and Y) using a
reference area as close to the center of the scene as possible. This enables us
to eliminate vertical misalignment and have a pivot point (the center of the
screen) to use for the rest of the steps (Figures 4.20 and 4.21) .