Questa 蠬a versione .html del file contenuto in http://www.cs.cmu.edu/~illah/PAPERS/firewire.pdf.
G o o g l e crea automaticamente la versione .html dei documenti durante la scansione del Web.
Per inserire un segnalibro o un collegamento alla pagina, utilizzare il seguente URL:http://www.google.com/search?q=cache:kF4Sf3cBZ_0J:www.cs.cmu.edu/~illah/PAPERS/firewire.pdf+pixel+jitter+digital&hl=it


Google non 蠣ollegato agli autori di questa pagina e non 蠲esponsabile del suo contenuto.
Sono state evidenziate le seguenti parole chiave:  pixel  jitter  digital 

Page 1
Advanced Imaging Magazine, January 2000, pp. 69-70
Firewire Untethered: High-Quality Images for Notebook Computers
Iwan Ulrich and Illah Nourbakhsh
The Robotics Institute, Carnegie Mellon University
5000 Forbes Avenue, Pittsburgh, PA 15213
iwan@ri.cmu.edu, illah@ri.cmu.edu
Until recently, no solution existed for acquiring high-quality color images in real-time with a
regular notebook computer, even though a high-quality color image acquisition system for
notebook computers would open the doors for a wide array of portable applications. In particular,
such a combination would be ideal for mobile robots that use high-resolution color vision, which
is part of our area of research.
While a large selection of frame grabbers is available for PCI desktop systems, only a few
have been available for PCMCIA notebook systems. Unfortunately, most PCMCIA frame
grabbers accept video only in the NTSC format, which allocates much more bandwidth to
luminance than color. Consequently, the NTSC signal is adequate for gray-scale image
processing, but its color content is too noisy for any reasonable color image processing. In
particular, the NTSC encoding allocates very little bandwidth to the blue channel, which is more
delicate than the red and green channels to begin with. The source of the problem is that most
CCD sensors are least sensitive in the blue spectrum. This problem is further exacerbated
indoors, where the illumination is often yellow and contains little power in the blue spectrum. As
a consequence, the already noisy blue signal from the CCD sensor is further weakened by NTSC,
resulting in a very poor signal-to-noise ratio.
Digital Cameras
In theory, higher-quality color images can be acquired from digital cameras than from analog
cameras. Digital cameras typically allocate the same number of bits for all three color channels,
thus avoiding additional encoding noise in the delicate blue channel. Of course, allocating equal
amounts of bandwidth to all three color channels is a property also found in analog systems that
use an RGB video format and frame grabber.
However, a digital system has two advantages over an analog RGB system: First, the A/D
conversion is performed close to the CCD sensor, thus keeping the amount of electronic noise to
an absolute minimum. After digitization, the video signal is immune to noise, such that a digital
camera can be several meters away from the computer without any risk of picking up additional
noise. Second, unlike analog camera systems, digital systems do not suffer from pixel jitter,
which is especially observable on the right side of images that have been captured with an analog
frame grabber. In fact, with a digital camera system, each captured pixel value corresponds to a
well-defined pixel on the CCD chip. Vertical edge detectors can thus be applied to digitally
captured images without any precautions for pixel jitter.

Page 2
Digital cameras that connect to a notebook computer through the serial port or the faster USB
port have been available for a few years. However, most of these cameras are intended for the
consumer market, and are equipped with cheap CMOS sensors to keep prices low. CMOS
sensors in general have much higher dark currents than CCD sensors. In fact, the quality of the
blue channel of most of these cheap consumer cameras is about as bad as that obtained with a
decent CCD camera and a NTSC frame grabber. Nevertheless, USB cameras offer a great low-
cost solution for acquiring images of NTSC-like quality. However, these cameras are limited to a
maximum throughput of only 12 MBits/s, resulting in rather slow frame rates for images of
decent resolution.
IEEE-1394 Digital Cameras
Fortunately, a new class of digital cameras that recently appeared on the market offers high
image quality as well as fast frame rates. These cameras, which adopt the IEEE-1394
specifications and communicate over the IEEE-1394 high performance serial bus, are an ideal
solution for acquiring high-quality images with a regular notebook computer. The fastest IEEE-
1394 cameras present image data at a guaranteed bandwidth of 400 MBits/s. This bandwidth
allows the acquisition of 30 frames per second, even at very high image resolution.
IEEE-1394 cameras offer other advantages beyond speed. Many IEEE-1394 cameras are
equipped with a high-quality CCD sensor, which substantially improves the quality of the blue
channel. IEEE-1394 cameras are also very versatile. The user can typically choose from several
modes of image resolutions and from several frame rates. Moreover, all adjustable camera
parameters can be controlled through the IEEE-1394 link.
For our research, we chose the DFW-V500 camera from Sony. The two photos show our set-
up, and the diagram illustrates the elements in it. The camera is equipped with a high-quality 1/3"
CCD sensor with progressive scan and square pixels. The camera supports four resolution
modes, with a maximum resolution of 640 x 480 in YUV (4:2:2) format. The frame rate can be
set at 3.75, 7.5, 15, and 30 fps. Basically all camera parameters can be adjusted through the
IEEE-1394 link: shutter speed, gain, gamma factor, brightness, sharpness, hue, saturation, and
white balance. Of particular note, the shutter speed can be adjusted from very short (1/100,000
sec) to very long (5 sec). This camera also adopts the popular C mount optical interface.
It is important not to confuse IEEE-1394 digital cameras with DV (digital video) cameras,
although both types communicate over the fast IEEE-1394 serial link. While IEEE-1394 digital
cameras are geared for the machine vision market, DV cameras are geared for the consumer
market. The main difference between the two types is image quality. DV cameras compress
image data in order to record longer movies onto a DV tape. Unfortunately, the image
compression is lossy. Although the image compression is optimized for human vision, the image
compression artifacts can pose serious problems for machine vision applications. In contrast,
IEEE-1394 digital cameras do not compress the image data, as the goal is maximum image
quality for each individual frame, not image storage.

Page 3
Software Support
The good news is that hardware support for notebook computers is available in the form of a
IEEE-1394 CardBus card manufactured by Ratoc Systems. The bad news is that until recently, no
software support has been available! While both the camera manufacturers and Ratoc Systems
claim that their products perform according to the IEEE-1394 digital camera specifications,
neither has provided the necessary software drivers for programmers to access the incoming
video stream.
Because we strongly believe in the future of these digital cameras, we chose to fill this void
by writing the software driver ourselves and making it available to the general public, at no cost,
on our web site: http://www.cs.cmu.edu/iwan/1394.
The result of our work is a C/C++ software library (DLL) that is compatible with Windows
98 and higher. The driver should also work under Windows NT, although it has not yet been
tested. In general, our software library can be used with any computer that is equipped with an
IEEE-1394 OHCI interface. In order to use our library with a regular notebook computer, the
notebook needs an open CardBus slot. The notebook can then be equipped with a low-cost IEEE-
1394 CardBus card, which provides the necessary IEEE-1394 OHCI interface.
In its current version, the software offers two different modes for acquiring images, and
allows control of all camera parameters according to the specific model. While the software has
only been thoroughly tested with the Sony DFW-V500 camera, it should work with other IEEE-
1394 models as well. Our web site also includes documentation, links, and the source code for a
simple demo program, which illustrates the use of the library routines.
External Power
8-30 VDC
IEEE-1394
Digital Camera
Sony DFW-V500
IEEE-1394
OHCI Interface
Ratoc REX-CBFW1
Laptop
Computer
Sony Vaio PCG-F270
C
a
r
d
B
u
s
1056 MBits/s
400 MBits/s

Page 4