Digital Signal Processing Reference
In-Depth Information
3
SAMPLING AND ALIASING
CHAPTER OUTLINE
3.1 Nyquist Sampling Rule
16
Whenever something is represented digitally, it must be
sampled. This means that for a continuous, unbroken object to be
digitized, it must be represented by many samples, close together
in either time or space. If the samples are close enough, then the
object appears as if it was continuous. Sampling in time, from
a video perspective, means that snapshots of the scene are taken
frequently enough to give an impression of continuously moving
video when played back. Use of 25 to 30 images per second is
common in video systems, and is generally acceptable to most
viewers (this will be double in interlaced video, but that will be
covered later).
Spatial sampling is applied to images, where enough pixels
must be used in both vertical and horizontal dimensions to
represent the image so that it appears realistic and sharp. The
sampling rates can change when an image is scaled, as this is the
process of converting from one resolution to another.
In order to take an analog signal and convert it to a digital
signal, we need to sample the signal. For a one dimensional
signal, we would use a device called an analog-to-digital
converter (ADC). A camera must sample in two dimensions
simultaneously, producing many samples at each instant. Unlike
traditional analog cameras that expose film using chemical
reactions to light, a digital camera uses a semiconductor device to
convert the light into electrical signals, which can be sampled by
many ADCs at the same time. A device called a charge coupled
device (CCD) is commonly used. A newer alternative is to use
complementary metal oxide semiconductor (CMOS) technology.
Such devices can take several million samples simultaneously
(this corresponds to the “Megapixel” rating used for digital
cameras), and convert them into electrical signals that represent
the individual pixels. These are then converted to digital repre-
sentation using ADCs.
