Graphics Reference
In-Depth Information
Finally, while this topic uses Hollywood movies as its motivation, not every visual
effects practitioner is working on a blockbuster film with a looming release date
and a rigid production pipeline. It's easier than ever for regular people to acquire and
manipulate their own high-quality digital images and video. For example, an amateur
filmmaker can now buy a simple green screen kit for a few hundred dollars, down-
load free programs for image manipulation (e.g., GIMP or IrfanView) and numerical
computation (e.g., Python or Octave), and use the algorithms described in this topic
to create compelling effects at home on a desktop computer.
1.2
THIS TOPIC'S ORGANIZATION
Each chapter in this topic covers a major topic in visual effects. In many cases, we
can deal with a video sequence as a series of “flat” 2D images, without reference to
the three-dimensional environment that produced them. However, some problems
require a more precise knowledge of where the elements in an image are located in a
3D environment. The topic begins with the topics for which 2D image processing is
sufficient, and moves to topics that require 3D understanding.
We begin with the pervasive problem of
image matting
— that is, the separation
of a foreground element from its background (Chapter
2
). The background could be
a blue or green screen, or it could be a real-world natural scene, which makes the
problemmuch harder. A visual effects artist may semiautomatically extract the fore-
ground froman image sequence using an algorithm for combining its color channels,
or the artist may have to manually outline the foreground element frame by frame.
In either case, we need to produce an
alpha matte
for the foreground element that
indicates the amount of transparency in challenging regions containing wisps of hair
or motion blur.
Next, we discussmany problems involving
image compositing and editing
, which
refer to the manipulation of a single image or the combination of multiple images
(Chapter
3
). In almost every frame of amovie, elements fromseveral different sources
need to bemerged seamlessly into the same final shot. Wires and rigging that support
stunt performers must be removed without leaving perceptible artifacts. Removing
a very large object may require the visual effects artist to create complex, realistic
texture that was never observed by any camera, but that moves undetectably along
with the real background. The aspect ratio or size of an image may also need to be
changed for some shots (for example, to view a wide-aspect ratio film on an HDTV or
mobile device).
We then turn our attention to the detection, description, and matching of
image
features
, which visual effects artists use to associate the same point in different views
of a scene (Chapter
4
). These features are usually corners or blobs of different sizes.
Our strategy for reliably finding and describing features depends on whether the
images are closely separated in space and time (such as adjacent frames of video
spaced a fraction of a second apart) or widely separated (such as “witness” cameras
that observe a set fromdifferent perspectives). Visual effects artists on amovie set also
commonly insert artificialmarkers into the environment that canbe easily recognized
in post-production.
