Database Compression Technologies - High-Performance Web Databases: Design, Development, and Deployment - page 710

MPEG can compress video at rates as high as 50 to 1. It has the added
advantage of allowing random access to video data and the synchroniza-
tion of video and audio tracks by multimedia workstations. Unlike JPEG,
MPEG takes an asymmetrical approach to compression (i.e., more comput-
ing power is required to compress full-motion video and audio than to
decompress it). Multimedia applications that require video to be com-
pressed only once and decompressed over and over again include training,
publishing, and reference databases, which are readily available on CD-
ROM disks.
There are some proprietary implementations of MPEG that improve
upon the performance of the industry standard, but which are not compat-
ible with it. For example, Cubic VideoComm, Inc., offers CVideoNow, which
it claims delivers a video stream comparable in quality to MPEG video at
1.5M bps over a connection of only 384K bps. CVideoNow Producer, the
utility for preparing video for streaming, uses proprietary compression
technology to achieve ratios of 500 to 1. Developers can exert fine control
over image quality during the encoding process. The video stream is eight-
bit color. Over a T1 or Ethernet connection, CVideoNow can deliver an
image of 320 by 240 pixels at around 15 to 30 frames per second.
Fractals
For databases that contain large image and graphics files that are not
intended for public access over the Web, there is a method of compression
known as fractal compression. This type of compression is used for amass-
ing large archives of image files used in the advertising and magazine pub-
lishing businesses, for example. It is also useful in engineering and design
applications. Its properties make it an effective format for certain types of
multimedia content intended for distribution over a corporate LAN.
With this type of compression, repetitive patterns on an image or
graphic file are identified and matched. Two patterns are considered the
same as long as one can be stretched, squeezed, or rotated into looking like
the other. The process continues until the image is the size specified by the
user. This process is computationally intensive, with compression gener-
ally taking longer than decompression (asymmetrical). Compression rates
of 100 to 1 are achievable with no loss of image quality. At 100 to 1 compres-
sion, 5M bytes of image information can be taken down to only 50K bytes.
Although JPEG products compress files quickly, their decompression
rate may be unacceptably slow for some applications (e.g., multimedia
training). With fractal compression, the reverse is true, making it suitable
for compressed databases that rely heavily on repeated decompressions
to server users.