Image compression eases storage, transmission issues
- By Pamela Houghtaling
- Jul 21, 1996
Electronic distribution of still images and video is becoming easier, and there is no doubt it will be a vital part of future agency communications and computing systems. But technical and cost barriers have to be overcome first, and key to this is the continuing advance in compression technology.
In some areas of government, this method of moving images is already a part of the daily activity. Law enforcement, the military and the burgeoning arena of telemedicine employ electronic image transmission extensively and are on the cutting edge of compression use.
Conceptually, compression is a simple process. Image data files are very big, and today's transmission pipelines are still relatively narrow. Compression slims down the image files enough that they can slide through the available pipes. But how you perform this weight loss, and what method you use, is important.
Compression schemes on the market today are either standards-based or proprietary. In the case of standards-based schemes, industry has assumed responsibility for product development, software, hardware and other equipment, all of which are widely available.
But if state-of-the-art, maximum compression is needed, then the customer will probably go with a proprietary, although more costly, scheme that was developed by a third-party firm, said Mark S. Bromley, director of core competency multimedia/imaging with Litton/PRC Inc.
Two of the best-supported standards-based approaches are JPEG and MPEG-2, which apply to still and video images respectively and which are used in a number of federal applications. Wavelet compression is probably the best-known of the emerging propriety technologies, and it is seeing use in the FBI.
JPEG stands for the Joint Photographic Experts Group, a committee within the International Standards Organization (ISO) that wrote the standard for still images. The JPEG standard works well for photographs of natural scenes and people, where some of the small detail can be lost without a noticeable degradation of the image. It has been adopted by the Defense Department as that agency's standard for still imagery.
"Today JPEG is the best because of the combination of quality, timeliness and flexibility of ratios," said Rafael Goldsmith, an official with DOD's Central Imagery Office, reporting on the results of a recent survey.
MPEG, the Moving Pictures Experts Group within the ISO, has developed standards for digital video and audio compression. Based on strong industry support and functionality, MPEG-2 was formally adopted last month as the standard for the U.S. defense and intelligence communities, said Anthony Galassi, a Central Imagery Office official.
Unlike the earlier MPEG-1 standard, MPEG-2 compresses the entire image, producing higher-resolution imagery for the viewer. As a result, however, MPEG-2 has greater communications requirements because the data files are bigger.
The Central Imagery Office is looking at commercial standards as a way of reducing systems maintenance costs, according to Goldsmith. Industry develops the system upgrades for standards-based products, which the customer can simply purchase. In the case of proprietary approaches, the customer takes on responsibility for maintaining the algorithm throughout the life of the applications, so costs are higher, PRC's Bromley said.
Interoperability is a critical issue for the military. Goldsmith cautioned, however, that standards do not in themselves guarantee interoperability because different vendors could implement only those segments that are in tune with their own business plans.
In order to guarantee that military users can exchange images, the Central Imagery Office has designed profiles of individual standards that specify requirements for each and has established a certification process for hardware and software products based on the requirements.
The Joint Combat Camera Center, a division of the American Forces Information Service, uses JPEG for all the still imagery it receives from its military photographers around the world.
"Image compression is paramount to us because they're transmitting the photographs to us over land lines or [International Maritime Satellite Organization] and telephone lines," said Master Sgt. Paul G. Robinson, the Joint Combat Camera Center's (JCCC) Webmaster and the technology and architecture noncommissioned officer to the American Forces Information Service's director of visual information. "We don't want huge files because we want to get the photographs today, not tomorrow."
JPEG is the only algorithm that allows compression high enough to achieve reasonable file sizes for transmission over telephone lines, he said. Robinson added that JPEG's range of compression ratios - anywhere from 10-to-1 to 100-to-1 - allows the user to select the quality required for the images.
For JCCC, the quality setting is in the good range at a compression ratio of about 15-to-1.
The use of commercial off-the-shelf software and equipment for image compression enables JCCC to provide photographs for dissemination to news organizations. "If we use proprietary compression routines, the public won't be able to access the photographs," Robinson explained.
The Armed Forces Radio and Television Service, meanwhile, is adopting the MPEG-2 standard for its satellite network. For Melvin Russell, director of AFRTS, the decision was clear-cut because the other compression choices were either proprietary or were standards covering still images, videoconferencing or industrial TV.
MPEG-2 provides for broadcast-quality video and enables the service to increase the amount of TV and radio that can be transmitted over the existing network to U.S. troops stationed around the world. When fully operational next year, the new MPEG-2-based system will deliver six broadcast-quality TV signals - instead of the single one available now - on the same satellite transponder at no increased cost.
Video compression enables organizations such as the AFRTS to increase service offerings using the same capacity on terrestrial or satellite networks, said Lisa Hobbs, marketing manager with Scientific-Atlanta's Satellite Television Networks Division, which is working on the service's project. Based on MPEG-2, Scientific Atlanta's PowerVu technology is making possible this expansion of capabilities for the service.
Conversely, Hobbs pointed out, if no additional services are needed, compression allows customers to cut back on their leasing costs by decreasing the network capacity required. Hobbs also noted that newer standards, such as Digital Video Broadcasting and MPEG-4, are already being developed, and these will offer even better methods of compression.
Proprietary compression techniques are being used in highly specialized government applications, such as fingerprinting. The scanned image of a fingerprint adds up to a file of over half a megabyte, according to Michael McCabe, a computer scientist with the National Institute of Standards and Technology's Information Access and User Interface Division. Using the wavelet scalar quantization algorithm, a fingerprint can be compressed into 37,500 bytes.
"It turns out that wavelet works very well with fingerprints," said Tom Hopper, a systems analyst with the FBI who has been involved with the development of the agency's Integrated Automated Fingerprint Identification System.
Wavelet offers greater efficiencies than other techniques such as JPEG, Hopper explained, and storage and transmission costs are substantially lower. A 15-to-1 compression ratio is currently being used for the fingerprint images - a ratio Hopper views as conservative. He believes that the ratio could even be larger without a loss of image quality.
With JPEG, the ratio could be only 10-to-1 or so before resulting in a loss of quality.
"Even small changes become objectionable when trying to make a difficult decision," he said.
A comparable JPEG image would require a much larger compressed byte size. JPEG compression, however, is being used for the Interstate Identification Index component of IAFIS.
"Wavelet is the near-term future of compression," according to Bernie Brower, an image scientist with Eastman Kodak Co.'s Commercial and Government Systems, who said a number of companies are developing wavelet-based techniques.
The Defense Advanced Research Projects Agency has been funding considerable research on wavelet, according to Anna Tsao, a program manager with DARPA's Applied and Computational Mathematics Program.
In digital video compression, such as that used in telemedicine and videoconferencing, proprietary methods are also finding use.
"In order to get the compression you need to run video over smaller links, such as the military has out to Bosnia, you need a higher compression ratio than is possible with a standard such as MPEG," said Michael Wassenberg, the technical marketing manager responsible for health care applications with GTE Government Systems.
GTE was chosen by the U.S. Army Medical Advanced Technology Management Office to provide network integration and field engineering support for the deployment of new telemedicine technologies for Operation Joint Endeavor in Bosnia.
Overall, the trend in compression technology is toward better image quality at higher compression ratios, decreasing the bandwidth and storage requirements and increasing the access speeds, whether to the desktop, a medical facility or a battlefield.
Standards-based methods of image compression are a less expensive way to go in view of the wide base of industry support and the availability of products.
Yet proprietary methods, such as wavelet-based algorithms, may be more appropriate for certain specialized applications requiring greater compression ratios, and they may become standards themselves over time.
In essence, user requirements play a pivotal role in determining the right method, industry and government experts said.
As NIST's McCabe advised, "First solve the problem."
Houghtaling is a consultant and writer based in Northern Virginia.