Graphics Reference
In-Depth Information
component function can be selected according to its properties.
Fractal compression
is based on the fact
that many signals are self-similar under scale. A section of the signal can be composed of transformed
copies of other sections of the signal. This is applied to rectangular blocks of the image. In fact, most
of the compression schemes break the image into blocks and then compress the blocks.
Codecs employ one or more of the techniques mentioned earlier. The names of some of the more pop-
ular codecs are Video I, RLE, GIF, Motion JPEG, MPEG, Cinepak, Sorenson, and Indeo 3.2. Microsoft
products areVideo I andRLE.Video I employsDCTcompression, andRLEuses run-length encoding. GIF
is an8-bit color image compression schemebasedon theLZWcompression. JPEGusesDCTcompression.
Motion JPEG (MJPEG) is simply the application of JPEG for still images applied to a series of
images. The compression/decompression is symmetric and is done in 1/30 of a second. JPEG compres-
sion can introduce some artifacts into some images with hard edges. The Moving Pictures Expert Group
(MPEG) standards were designed specifically for digital video. MPEG uses the same algorithms as
JPEG for intraframe compression of individual frames, called I-frames. MPEG then uses interframe
compression to create B (bidirectional) and P (predicted) frames. MPEG allows quality settings to spec-
ify the amount of compression to use [
24
] and can be set individually for each frame type (I, P, B).
MPEG-1 is designed for high-quality CD video, MPEG-2 is designed for high-quality DVD video,
and MPEG-4 is designed for low-bandwidth Web applications [
22
].
Cinepak and Sorenson are products initially targeted for the MAC world, although Cinepak is now
available for the PC. Cinepak uses block-oriented vector quantization. Sorenson uses YUV compres-
sion with 4
4 blocks and employs interframe compression [
20
]. Indeo 3.2 is an Intel product that also
uses block-oriented vector quantization. Cinepak and Indeo are highly asymmetric, requiring on the
order of 300 times longer for compression than for their efficient decompression. Indeo also incorpo-
rates color blending and run-length encoding into its scheme.
Digital video formats
The codec products (as opposed to the underlying codec techniques) used for DV have an associated file
format. Some formats also include timing information, the ability to animate overlay images (sprites), and
the ability to loop over a series of frames. MPEG and MJPEG are both common DV formats. GIF89a-
based animation (animated GIF) is basically a number of GIF images stored in one file with interframe
timing information but no interframe compression. Compressing continuous-tone images can result in
color banding. The compression does well on line drawings but not on complex outdoor scenes. GIF
animations can use delta frames, which, for example, overlay images on a previously transmitted back-
ground. This saves retransmitting static information for some animations. As used here,
movie format
refers to a format that is codec independent and that can handle audio as well as imagery. Both Quicktime
codec can be used with these standards as long as a compatible plug-in is available. Cinepak has been a
standard codec used with Quicktime, but Quicktime can also accommodate other codecs such as Sor-
enson and JPEG [
24
]. Quicktime organizes data into tracks and includes timing information. Video for
Windows from Microsoft uses Video I and RLE as standard codecs but can also accommodate others.
Video for Windows allows interleaving of image and audio information.
Digital television formats
Most of the DTV formats are based on sampling scaled versions of the color difference signals (
B-Y
,
R-Y
). Luminance and the scaled color difference signals are referred to as YUV. Common formats are
D1, D2, D3, D5, D6, Digital Betacam, Ampex DCT, Digital8, DV, DVCam, and DVCPRO. When
Search WWH ::
Custom Search