Videoconferencing
Improved
communications infrastructure and affordable PC-based systems make
videoconferencing a viable mass market option
By Simon Burns
At
the 1964 World’s Fair in
The
Picturephone joined history’s long list of
fascinating but unattainable ideas, such as pocket televisions and personal
computers. The true dawn of the videophone age would have to await development of a communications infrastructure able to carry the huge
volumes of data demanded by video transmissions. Today, new digital
communications networks and advances in data compression technology are finally
making mass production of videophones and videoconferencing application a
possibility. Meanwhile, worldwide standards are being finalized systems from
different manufactures will be interoperable.
Technical
issues
Since the earliest days of videoconferencing, the biggest
stumbling block has been the inadequate bandwidth available on public
communications systems . Bandwidth refers to the capacity of a channel, a
telephone line, for example, to carry information. In videoconferencing
applications, bandwidth is generally measured in kilobits per second (Kbps) or
megabits per second (Mbps). Given an ideal communication channel, how much
bandwidth would videoconferencing system required? The CIF (Common Intermediate
Format) size, supported by the majority of videoconferencing application, is
253 x 288 pixels. This is less than a quarter of the screen on a 15-inch
monitor. At eight bits per pixel, each frame of video will occupy about 811,000
bits. Assuming we want something close to TV quality, we need to send 25 frames
per second. This gives us a total of 25 x 811,000 or more than 20 million bits
per second.
Since videoconferencing requires twoway
communication, this figure should be doubled , to 40
Mbps . For comparison, the fastest modems can transmit at up to 33.6 Kbps, over
1200 times slow. Even an Ethernet LAN , at 10 Mbps ,
will be found wanting . These figures illustrate the importance compression
to current videoconferencing schemes .
Compression algorithms, such as MPEG can reduce a stream of video data
to less then ten percent of its original size. While MPEG is suitable for the
compression of fast moving images of the type found in a feature film, the
subject mutter of most videoconferencing sessions presents a slightly different
challenge.
Although basically
similar to MPEG the H.261 compression algorithm defined under the International
Telecommunication Union(ITU) H.320 videoconferencing standard is optimized for
«talking heads» scenes with little motion . More specifically, H.261 is at its
best when handling motion vectors of less than 15 pixels per frame. In practice,
this means that the head and facial movements present in normal conversation
are most suitable for transmission by H.261. Rapid motion, waving for example,
is likely to lead to a temporary loss of video quality.
Another area where H.261 differs from MPEG, is
its need to operate in real time on unreliable networks. Both compression
algorithms achieve a considerable part of their bandwidth savings by only
transmitting the parts of the picture that change from frame to frame. When
data is lost during transmission - a strong possibility when video is
being sent down a telephone line - the following frames will be garbled. To
recover quickly from this situations, H.261 reqularly sends a complete frame to video data.
Because there are strong similarities between H.261 and MPEG (they are
both based on a compression scheme known as Discrete Cosine Transform, or DCT),
it would seem logical to produce hardware able to handle both. Some companies
such as Information Technology Inc. (ITI) of
Communication
channels
While there is a very wide variety of software and hardware in use for
desktop videoconferencing systems, there are only three widely accepted methods
of connecting desktop to desktop. These are ISDN (Integrated Services Digital
Network), LAN and POTS (the Plain Old Telephone Service).
POTS systems are particularly attractive to buyers because of the
ubiquity, familiarity and low cost of the standard analog phone line. The
problem with POTS, however, is its limited bandwidth. State-of-the-art 28.8
Kbps modems are, on a good line, fast enough to transmit a small (credit-card
sized) image at around ten frames per second.
If
there were faster modems on the horizon, this would be an acceptable beginning.
But 33.6 Kbps modems recently announced by manufactures such as US Robotics, are coming very cloe to
the theoretical maximum amount of data that can be squeezed into the narrow
pipe of an analog telephone line. Any further improvement in POTS
videoconferencing will have to come from enhanced compression algorithms.
Integrated Services Digital Network (ISDN) , a
key enabling technology for videoconferencing has been a long time coming .
Pilot ISDN projects were established in the 1970s. Today there are an estimated
300,000 ISDN lines in use in
Photo-journalism is another strong potential market for today’s cameras.
In fact one news agency, the Associated Press, even markets a re-badged Canon digital camera, the NC-2000, to journalists.
Relatively low resolutions and extremely large sizes are still hindering
acceptance in this field, however.
Consumer electronics manufacture Casio has something of a reputation for
quirky products that use design innovations to undercit
competitors on price or features. This has always been a
rather hit or miss approach, but Casio’s first digital camera looks
likely to be a hit.
Most of the digital cameras on the market still resemble film cameras
very closely in appearance and operation. Casio’s QV-10 moves away from this
traditional analogy by replacing the viewfinder with a 1.8-inch, 480x240 active
matrix LCD screen of the sort familiar from some video
cameras. Updated about three times a second, the LCD provides an accurate
preview of the picture to be taken and can be used to review shots stored in
the camera’s 96-picture memory. The side-mounted lens, swivels vertically
though 180 degrees, allowing users to keep the subject in shot and watch the
LCD, while the camera is held at waist-level or above the head, for example.
The
camera also features TV output, enabling direct connection to a TV, video
recorder, or Casio’s US$600 VG-100 Personal Video Printer. As well as a digital
output socket, used to download images to a PC, the QV-10 also has a digital
input. Images can be uploaded from a PC to the camera, and the built-in LCD or
any TV screen can then be used for presentations.
The QV-10’s unusual design combined with the obvious utility of the LSD
screen will make it stand out on store shelves that are already becoming
crowded with sub-$1000 digital cameras featured here, the QV-10 possesses an
external feature, the LSD screen, that makes demonstrations simple and
immediately shows a potential buyer why a digital camera is different from, and
perhaps better than, a traditional camera.
A
group of sub-1000$ cameras from several manufactures, including Kodak, Chinon, Logitech and Apple, are based around a core
developed by Kodak and Chinon, a Japanese
manufacturer products.
The Kodak DC-40 and the Logitech Pixtura are
practically twins. Designed by Logitech and engineered by Kodak, the two
cameras are physically identical, differing only in some minor detaiils of their internal operating software. The
description of the Pixtura, below, also applies in
most respects to the DC-40.
Logitech’s Pixtura is a little larger than the
kind of fixed-focus point and shoot camera that its manufacturers hope it will
supercede. Attractively-styled, it takes 24-bit color pictures at a maximum
resolution of 768 by 512 pixels. The Kodak DC-40 differs slightly here with an
upper limit of 756 by 504 pixels. The Pixtura
stores up to 48 shots at the highest resolution, and up to 144 at lower
resolutions.
One problem afflicting all four of the sub-$1000 cameras mentioned here
is the appearance of spurious light or dark pixels, sometimes referred to as
artifacts, on high-contrast boundaries within the image. Some sources attribute
this problem to the CCD, others blame Kodak’s proprietary RADC algorithm, which
compresses data inside the camera.
Chinon’s, ES-3000, while based on the same hardware as the Pixtura, adds additional features and is aimed at a
slightly higher market segment. The ES-3000 turns out the same 24-bit 640 by
480 pixel pictures but adds features such as a 3X power zoom lens and auto
focus. Chinon’s recommended price for the unit is
below US$1000. Another Chinon camera, the ES-1000, at
US$499, is probably the first color digital camera to break the US$500 barrier.
However, resolution is only 501 by 370 pixels, and the camera only stores eight
images.
There is a large gap in the market between the best of the
VGA-resolution cameras, at between US$1000 and US$3000, and the more powerful
cameras, which cost at least US$10,000. Pitched some way above the cheaper
cameras are models such as Agfa’s ActionCam,
which provides resolution of 1528 by 1148. (Contact Agfa
Division, Bayer Corporation, at Tal + 1 800 685 4271,
Fax +1 508 583 4168.) In the same league is
For those who want quality at any price, Kodak’s DCS-460 offers a
resolution of 3060 x 2036 at a recommended price of $27,995. The most expensive
model on the market is probably the Big Shot, from US imaging company, Dicomed. The ability to take 4096 by 4096 pixel shots in
24-bit color will leave very little change from US$50,000.
Most of the advice given to users of conventional cameras, regarding exposure and picture composition, is also relevant to digital cameras. To use a digital camera to best advantage, though, some other points should be borne in mind. A digital camera’s CCD is more sensitive to light than photographic film. This fact is appreciated by astronomers, who, after commercial video and photography companies, are one of the main users of CCDs. Photographers used to traditional cameras may find themselves overexposing shots at first. In some of the cameras, the CCD is equivalent to ISO 200 film, while the more sensitive CCDs, in cameras such as the Agfa ActionCam, reach ISO 800, or even ISO 1600. In practical terms, this means that pictures can be taken in relatively low light without a flash. In general, it is better to underexpose shots and then use image processing software, as necessary, to brighten them up.
The shutter speeds of some of the cheaper cameras are inadequate for
shots of faster moving objects, which will appear blurred. Several of the
cameras mentioned here suffer from aliasing or artifacting
problems. Pictures with many high-contrast edges, a checker-board pattern or
ivy growing on a white wall, for example, will bring out the worst in these
models.
Familiarity with a good image processing application, Adobe Photoshop or
Fractal Design Painter for example, will pay dividends with the low-end
cameras. Many users may find themselves spending more time on post-processing
work than on taking pictures. Many business users require a close-up or macro
lens, particularly for applications where a digital camera is replacing a
flatbed scanner. Some of the models mentioned here have this feature built in,
others can be enhanced with a proprietary or third-party lens.
The video applications that have traditionally driven CCD development have been satisfied with relatively low resolutions, but even casual users of digital cameras will require better quality from their snaps. The resolution of a typical frame of 35mm photographic film is 6,000 x 10,000 pixels. The VGA (640 x 480 pixel) resolutions available from today’s sub-$1,000 digital cameras fall woefully short of these figures. Like LCD panels, CCDs fase size, restrictions related to manufacturing difficulties. There is an increasing likelihood of defects being introduced as the number of elements increases, so larger CCDs produce higher reject rates, which leads directly to increased unit prices. CCDs are particularly vulnerable to defects because of the way in which an image on the CCD is transferred to the camera’s memory. There is a fairly high probability that even a single fault will make the whole CCD useless.
The hugely cletailed images captured by
even higher resolution cameras will require cheaper storage and better compression
algorithms. Today’s comparatively small CCDs are
already creating some very large files, up to 20 MB in extreme cases. In
digital form, the 60 million pixels of a single 35 mm film frame, stored as a
24-bit true color image, would occupy around 120 MB of storage space, without
compression. Disk storage offers cost benefits over RAM but falls down when
durability is required, especially outside the studio. However, lenses have
always been fragile, so photographers are used to handling expensive camera
equipment with a reasonable degree of care.
The home market will create a great demand for cheaper color printers.
In the case of high-end printers, particularly dye sublimation models, both the
machines themselves, and the print media and other consumables, are still much
too expensive for ordinary weekend photographers. There are interesting
implications here for the printer trade in general. If even a part of the
billions of dollars that are spent on film each year is diverted to the printer
market, production volumes will rise, and prices will fall so low that
monochrome printers will, at some point, go the way of monochrome monitors.
Kodak CEO, George Fisher has postulated a US$300 photographic quality color
printer, to complement the future US$300 digital camera.
