| Anatomy
of a Digital Image: bits, bytes, and pixels
What is the Difference between digital and
analog images?
First, let us establish a feel for the analog
or digital process by analogy:
Real numbers are analog........Integers
are digital
Horseshoes is analog.............Darts is digital
A dimmer switch is analog......A standard light switch is digital
Mittens are analog................Gloves are digital
Analog phenomenon are continuously variable
typical of real world events. A digital signal is an approximation
of the analog "real world" signal. Ultrasound signals
are analog however they are converted to a digital format
early in the endosonography machine. Analog or continuous
signals are converted to digital by regularly sampling the
signal and converting its value into a discrete number. The
more frequently we sample the phenomenon, the better we can
approximate it with our digital representation. Digital advantages
include less expensive processing and manipulation and no
loss of signal (or binary loss of signal) on transmission,
archive and reuse. The major disadvantage is that the signal
is an approximation to the "real world" or analog
phenomenon. This disadvantage can be negligible with proper
sampling.
Since ultrasound images are formed and processed
digitally, they can be archived digitally (on tape, CD, DVD
etc.), analyzed digitally (digital computer) and communicated
digitally (from a digitally driven flat panel display or Internet).
They might also be archived in an analog format (VHS, SVHS
tape or printer output), or communicated (from an analog CRT
or across a modem) and then re-digitized for further display,
distribution or analysis. This analog back to digital conversion
introduces artifact and degrades resolution by transitioning
back and forth and is often the culprit in poor image quality
in medical image publications today. This publication is digital
and we encourage the submission of images that have remained
digital since their formation.
In terms of a two-dimensional endosonography
image, one axis is the distance of an ultrasound reflector
from the transducer. The other axis indicates the direction
of the transducer in a plane swept by rotation of the transducer.
These two axes define an image plane and distances can be
measured on these two axes with an accuracy of about one part
in 500. Consequently the two axes generally define a plane
space of 512 x 512 sampled points. Each sample point is a
pixel and is regularly spaced (digital).The image content
(value stored in a pixel) is the intensity of ultrasound detected
from each point in the measurement plane. If we magnify a
digital image we can eventually see the domain of each pixel(Figure
1). If we don't magnify and have sufficient resolution then
we may not be able to decipher whether the image is digital
or analog.
In the early days of ultrasound, the detectors
indicated ultrasound above a certain threshold, i.e., the
ultrasound was either present or absent. This is the basis
for forming a binary image, an image with pixels that are
either on or off, an image with one bit deep pixels. On or
off can be attributed to any physical indicator such as white
for on and black for off (Figure 2)
As the ultrasound measurement instrumentation
becomes more sensitive, small differences in intensity can
be detected. Ultrasound intensity might be split into four
categories such as: strong, mild, weak, and absent and therefore
representable with two bits or two on/off switches, 00, 01,
10 and 11. This corresponds to a two bit deep pixel in the
image. Today, ultrasound intensity is measured to an accuracy
of about one half to one per cent, an accuracy no better than
one part in 256. 256 is 28 and therefore, eight binary switches
or an 8 bit deep pixel is used for each storage location in
the image(Figure 3).
Eight bits is referred to as a byte and
it is economically feasible today to manipulate and store
data in bytes. Therefore, future improvements in the ultrasound
detection accuracy, for example a factor of two improvement,
may see a jump to two byte or sixteen bit deep pixels instead
of the required nine bits to store a factor of two improvement.
With 8 bit deep images we can use each pixel value to generate
brightness on a display or the quantity of ink on paper etc.
Again, a common convention, but by no means necessary, is
to represent the pixel value of 0 as black, the pixel value
of 255 as white and the values between as linearly varying
shades of gray.
If the image display is capable of color
(Figure 4)then we have to make decisions about the colors
represented by each of the values stored in the pixels. There
are many different schemes for assigning colors to possible
pixel values and some of these will be covered in a future
column.
References
1. Moving Theory into Practice, Digital
Imaging Tutorial. Cornell University Library. http://www.library.cornell.edu/preservation/tutorial/
2. Kruenen Ben. Big Ben's Digital Imaging Tutorial. http://www.bigbenpublishing.com.au/digital/
3. Bouton, DG, Bouton, B, Kubicek, G, Nathanson, MZ, Rich
J, Ward A. Inside Adobe Photoshop 6, Limited Edition,
2001, Bouton, DG (editor). New Riders Publishing,
Indianapolis, IN.
|
