Brinkmann R. The Art and Science of Digital Compositing
Подождите немного. Документ загружается.
292 The Art and Science of Digital Compositing
feature might include. Don’t use this as a ‘‘yes or no’’ checklist, but instead try
to examine the features of the package in question to understand how complete
the implementation might be.
Finally, remember the importance of the overall design of the package’s user
interface. This is probably the most difficult component of a package to evaluate,
since it requires spending time with the product to understand the particular
paradigm. It is something that cannot be dealt with in a simple features list, but
absolutely should not be ignored. Chapters 8 and 9 go into greater detail about
certain features and methodologies that are relevant to this topic.
The categorization of the features in this list is rather arbitrary, as is the naming
of the tools themselves. Some of the features described below are named fairly
consistently throughout the industry. For instance, ‘‘Brightness’’ almost univer-
sally refers to a tool that multiplies all three channels of an image by a constant
value. In situations such as this, we’ll give that name for the feature before we
describe it. But many features can be referred to by completely different terminol-
ogy depending on the software you are using. If you are using this list to help
evaluate the functionality of a particular piece of software, you may need to
explore a little bit before you can accurately determine whether, and to what
extent, the software in question supports a specific feature. Remember too that
many operators can be built as a combination of other operators. The list here
consists primarily of tools that either cannot be built from other tools or are
common enough to warrant inclusion.
COLOR CORRECTION
Although we will describe many of these tools in terms of the operations they
perform on each pixel, in practice most of these tools should also be able to perform
individually distinct operations on each channel of an image as well.
Add: Add a constant value to each pixel in an image.
Brightness: Multiply the RGB channels by an equal, specific amount.
Clamp: Clamp the value of each pixel in an image to be between a specific
range. Values outside of this range will be forced to the limits of the new range.
Color-Space Conversions: The ability to convert images between various color
spaces. These include RGB, HSV, HLS, CMY, CMYK, as well as a number of
custom color spaces (such as the Cineon color space commonly used in film
work).
Compress: Compress the value of each pixel in an image to be between a
specific range. The entire gamut of values between 0 and 1 is scaled to fit
within this new range.
Digital Compositing Software: Tools and Features 293
Contrast: Modify the range/ratio of brightness values in an image. The mid-
point of this contrast should be something that the user can control.
Divide: Divide the value of each pixel in an image by a constant value.
Expand: Choose a specific range of pixel values and expand them to fall in the
range of 0 to 1. Values that were originally outside the range chosen will be
pushed to 0 and 1 as well.
Expression Language: A robust language that can be used to modify the colors
in an image based on arbitrary mathematical expressions. Ideally, this language
will be capable of using information from more than one source image.
Fade: Multiply the RGBA values of a four-channel image by an equal, specific
amount.
Gamma: Modify the apparent brightness of an image by raising or lowering
the midrange values of the image.
Histogram Equalization: Stretch the range of values in an image to fit as
completely as possible between 0 and 1.
HSV Modification: Modify an image based on its hue, saturation, and value.
Invert: Invert the value of each pixel in an image, producing a negative of the
image.
Look-up Table: Apply a user-defined table that specifies a particular mapping
of input pixel values to output values.
Matte Divide: Divide the RGB channels of an image by its alpha channel.
Matte Multiply: Multiply the RGB channels of an image by its alpha channel.
Multiply: Multiply the value of each pixel in an image by a constant value.
Reorder: Shuffle the order of the channels in an image.
Saturation: Modify the saturation of an image.
Set: Set any channel or set of channels to a specific constant value.
SPATIAL FILTERING
Blur: Apply some type of blurring algorithm. The user should be able to choose
different horizontal and vertical blur values.
Convolve: Convolve an image with an arbitrary filter. The size of the kernel
should be user-definable as well.
Dilate: Increase the coverage of bright areas in the image, while decreasing
the dark areas.
Emboss: Modify an image to give it the sense of additional depth. Usually
based on the luminance values in the source image.
294 The Art and Science of Digital Compositing
Erode: Decrease the coverage of bright areas in the image, while increasing the
dark areas.
Median: Replace the value of each pixel in an image by the median value of
the neighboring pixels.
Sharpen: Apply some type of sharpening algorithm. The user should be able
to choose different horizontal and vertical sharpening values.
GEOMETRIC TRANSFORMATION
For geometric transformations, you should be able to control whether each event
occurs within a given resolution or whether it produces an image with a new
resolution. For instance, a rotation applied to a specific image could either crop
any piece of the image that rotates outside the original frame size, or it could
cause the output resolution to be increased to prevent any such cropping. Portions
of an image that are moved out of frame by a geometric transformation should
not be permanently cropped. Instead, the data should be available for use by
subsequent transformations. Optionally, you should also be able to specify whether
you wish any portion of the image that moves out of frame to wrap around and
reenter the frame on the opposite side.
The user should be able to choose the type of filtering that is used for most
transformations. There are times when one may wish to choose different filtering
along the vertical and horizontal axes.
Finally, for any animated geometric transformation, you should be able to
choose whether or not the result is rendered with motion blur, and be able to
control the extent of the motion blur that is applied.
Crop: Remove a portion of an image that lies outside a specific boundary, either
by specifying two corners that define a rectangular region or by specifying a
single point and the resolution of the desired result.
Fit: Scale an image to fit within a given resolution, usually with the option of
maintaining the aspect ratio of the original image.
Lens: Create or undo the distortion caused by a camera’s lens.
Pan: Reposition an image within a given frame.
Pin: Distort an image by moving four arbitrary points within the image to four
new positions.
Resize: Change an image’s size by specifying a new resolution.
Rotate: Rotate an image by specifying the center of rotation and the amount
of rotation.
Scale: Scale an image by giving a multiplier for the X and Y axes.
Digital Compositing Software: Tools and Features 295
Shear: Shear an image around a given point, specifying shear amounts for both
the X and Y axes.
Twirl: Distort an image by twirling it around a certain point. You should be
able to control the fall-off on the twirl as a function of distance from the center
point.
Warp: Warp an image based on a control grid or a set of control splines.
Expression Language: A robust language that can be used to warp an image
based on arbitrary mathematical expressions. Ideally, this language will be
capable of using information from more than one source image.
Flip: Mirror the image around the X axis.
Flop: Mirror the image around the Y axis.
MISCELLANEOUS
Grain: Add simulated film grain to an image. You should either be able to
specify a variety of parameters that describe grain characteristics (such as grain
size, density, luminance and chrominance variance, etc.) or to choose from a
list of different film stocks that you wish to emulate.
Radial Blur: Blur or smear an image radially about a certain point. You should
be able to control the fall-off on the smear as a function of distance from the
center point.
Smear: Smear an image in a certain direction by a certain amount. Often used
to simulate motion blur.
IMAGE COMBINATION
The image-combination tools are described using the same conventions that were
discussed in Chapter 4. Thus, A and B are considered the two source images. If
there is a need to specify certain channels in an image explicitly, they will be
indicated with subscripts; otherwise it is assumed that the operation will be
performed on all four channels of an image equally. All images are assumed to
be four-channel premultiplied images.
Add: Add two images together on a pixel-by-pixel basis (A Ⳮ B).
Mix: Combine two images by using a weighted pixel average. (MV ⳯ A) Ⳮ
[(1 ⳮ MV) ⳯ B], where MV is the mixing value.
Atop: Place the foreground over the background, but only inside the back-
ground alpha. Thus, A atop B is really the same as (A in B
A
) over B.
Reorder: Move specific channels between two different images.
296 The Art and Science of Digital Compositing
Displace: Warp an image by using a second image to control the amount of
warp at any given location.
In: Retain the foreground only within the background’s matte. Thus, A in B is
the same as A ⳯ B
A
.
Luminosity: A tool to modify the matte channel of a four-channel premultiplied
image. Decreasing the matte channel gives the image a luminous look when
it is placed over a background. Also known as the Opacity operator.
Max: Use the maximum value of the equivalent pixel location in the two images.
Min: Use the minimum value of the equivalent pixel location in the two images.
Multiply: Multiply one image by another image (A ⳯ B).
Out: Retain the foreground only outside of the background’s matte. Thus, A
out B (or ‘‘A held out by B’’) is the same as A ⳯ (1 ⳮ B
A
).
Opacity: See Luminosity.
Over: Place the foreground over the background using the foreground alpha.
A over B is the same as A Ⳮ [B ⳯ (1 ⳮ A
A
)].
Screen: Invert both images, multiply them together, and then invert the result.
1-[(1 - A) ⳯ (1 - B)].
Subtract: Subtract one image from another image (A ⳮ B). You should be able
to choose whether the value of the resulting pixel is clipped to 0 if the result
is a negative number or if the value of the resulting pixel becomes the absolute
value of the result.
Under: Place the background under the foreground using the foreground alpha.
Thus, A under B is the same as B over A.
Xor: Retain both source images only where their mattes do not overlap. A ⳯
(1-B
A
) Ⳮ B ⳯ (1-A
A
).
Z Compose: Choose the resulting pixel from either source image by comparing
the Z-depth values for each image.
CONTROL
You should be able to control and limit how a given tool is applied. Various
control methods include:
• The ability to restrict the operation so that it only occurs on specific channels
of the image.
• The ability to restrict the operation so that it only occurs on the even-num-
bered or the odd-numbered scan lines of the image.
Digital Compositing Software: Tools and Features 297
• The ability to specify a rectangular region of interest (ROI) outside of which
the effect will not be applied.
• The ability to control the amount of the effect that will be applied, based on
the grayscale values in an additional (separate or integrated) matte/mask
image.
• The ability to limit the effect so that it will only be applied if it changes a
given pixel by a user-specified threshold.
You should also be able to have some control over how the software uses the
resources of the system on which it is running. This includes the ability to limit
the amount of memory that the software will use, as well as having control over
what disk space is used for any temporary image storage.
FIELD CONTROLS
Interlace: Merge two images into a new image by choosing alternating lines
from each image.
Deinterlace: Separate a single image into two new images, one containing only
the even-numbered lines from the original image and one containing only the
odd-numbered lines.
Swapfields: Switch the even- and odd-numbered lines in an image.
MATTE GENERATION
Lumakey: Create a matte based on the luminance values in an image.
Chromakey: Create a matte based on a specific thresholded color in an image.
Difference Matting: Create a matte based on the differences between two im-
ages.
Specialized Tools: Any of a number of proprietary tools for creating a matte
for an object. These tools may rely on the fact that the object was shot in front
of a uniformly colored background or may use other techniques, such as motion
analysis, to determine the matte’s location and transparency.
TIMING/DURATION
Append: Append one image sequence to another to create a new, longer se-
quence.
Drop Frames: Remove frames from a sequence at regular intervals, either by
specifying which frames to keep or which frames to drop.
Duplicate Frames: Duplicate certain frames in a sequence at regular intervals.
298 The Art and Science of Digital Compositing
Hold Frame: Convert a single frame into a sequence of identical frames.
Pulldown: Perform the standard conversion between 24-fps film and 30-fps
NTSC video.
Pullup: Perform the standard conversion between 30-fps NTSC video and 24-
fps film.
Reverse: Reverse the order of the frames in a sequence.
Time Compress/Expand: A tool for shortening or lengthening a sequence of
images, either by selectively averaging the frames in a sequence together (usu-
ally with some kind of weighted mix) or via the use of certain specialized
motion analysis algorithms.
IMAGE GENERATION
Constant: Create a constant-color image.
Gradient: Create an image that contains a horizontal, vertical, or radial gradient
between different colors.
Paint Tools: A broad category of possible features. An ideal group of paint
tools would allow for procedural painting that can be applied to a sequence
of images instead of just the one in question.
Shape Drawing: Create an image that contains user-defined shapes created via
curves or polygons. The user should be able to control the color and transpar-
ency of the shapes and the softness of the edges.
Text: Create text with a user-specified font, color, size, orientation, and place-
ment.
TRACKING
There are a number of features that would be considered to be related to tracking.
The following lists a few specific items, and Chapter 7 provides much more
detailed information.
Subpixel Accuracy: The ability to accurately track features at greater than whole-
pixel resolutions.
Multiple Point: The ability to track several points at the same time.
Integration: Tracking information should be easily applicable to most geometric
transformation tools, including Pan, Rotate, Scale, Pin, and Warp.
Digital Compositing Software: Tools and Features 299
Stabilization: Choose a specific point and stabilize the plate based on the as-
sumption that the point in question is unmoving.
OTHER
Finally, there are a number of additional features that most compositing packages
should support. These include the following:
Resolution Independence: The ability to translate compositing parameters that
have been created to work on a sequence of images at a certain resolution to
work with a sequence at a different resolution.
File Format Independence: Support for reading and writing a wide variety of
image file formats, such as those listed in Appendix C.
Platform Independence: The ability of a given piece of software to run on a
variety of different hardware platforms, often from different manufacturers.
Extensibility: The ability to add functionality to the package via the use of
macros, plug-ins, and so forth.
Optimization: Automated tools that can analyze and optimize complex com-
positing scripts.
Batch Processing: The ability to create complex scripts that define large hierar-
chical compositing processes.
Internal Accuracy: The ability to process data at different bit depths. Common
depths include 8 and 16 bits per channel, as well as support for a floating-
point mode, which would maintain pixel values that are pushed beyond the
range of 0 to 1.
Proxy Support: The ability to use low-resolution stand-ins for the image/
sequence in question, and the ability to prototype effects or scripts based on
these proxies.
This Page Intentionally Left Blank
A
PPENDIX
B
Digital Compositing
Software:
Manufacturers
The following is a list of companies that provide software tools for the creation
and manipulation of digital images. While we have concentrated on companies
that produce digital compositing software, most of these vendors also provide a
variety of other tools, including 3D animation and rendering packages, digital
editing systems, morphing software, and digital paint programs. Please consult
the individual companies’ Web sites for further details about the specific markets
that they address.
In any list of this sort, there will inevitably be inaccuracies and exclusions, but
at the time this book was written, the list was a reasonable overview of the major
providers in the industry. Obviously, both specific products and the companies
that provide them can change names or disappear entirely, and given the volatile
nature of the software business, this list may actually be changing more rapidly
than the technology itself.
Adobe Systems, Inc. www.adobe.com
Advanced Digital Imaging, Inc. www.adii.com
Alias Wavefront www.aw.sgi.com
Avid Technology, Inc. www.avid.com
Chyron Corporation www.chyron.com
Composite Components Company www.digitalgreenscreen.com
301