mmo-ga:
You're right that there are a lot of technical definitions out there;
since you asked for a lay-person's description of what interpolation
really means, let me explain it in my own words instead.
We'll start with the digital camera primer from "How Stuff Works":
http://electronics.howstuffworks.com/digital-camera11.htm
On this page (the 11th page of the primer), it briefly mentions that:
"This process of looking at the other pixels in the neighborhood of a
sensor and making an educated guess is called interpolation."
So how does it really work?
Well, think of the sensor in a digital camera as being like you
standing behind a screen-door (you know, the mesh kind), and looking
outside. Each square of the mesh represents one pixel. Each pixel has
a different value for color and brightness. All of the squares in the
mesh (ie. the entire screen door) combine to form the actual picture.
Remember that each square, or pixel, can only have one value for color
and one for brightness, though. So, if you took a smaller part of this
screen, and stretched it to double its size, with each square still
having only one color and one brightness, now all of a sudden your
picture is rather fuzzy and made up of these big squares. This is
where interpolation comes in. The computer in the camera looks at each
of these bigger squares, and tries to divide it into four smaller
squares. If it did this by just assigning each smaller square the same
color and brightness as the original bigger square, though, your
picture is still going to be fuzzy. So, what the computer does is it
steps back (figuratively speaking), and looks at what's going on in
the other big squares around the one it's trying to divide up. It
tries to smooth the color and brightness changes so that the end
result is that the whole picture looks smooth and less fuzzy.
The problem with interpolation, though, is that it is all just
guess-work by the computer. Since the value of each square is a single
color and brightness, what happens if there is a piece of detail that
happened to fall exactly between two of the original squares, and was
not captured by the camera? Well, that detail gets lost, since
interpolation can only guess at what "should be" between the original
squares.
Is interpolation only used in the situation you described? Actually,
no. All digital cameras use a form of interpolation to function. You
see, in the sensor in the camera, you actually have a grid of little
individual sensors, each capable of sensing only one of either red,
green, or blue. When you click the shutter, all of these sensors are
simultaneously exposed to "the picture" that you're trying to take.
Each sensor picks up a certain level of brightness in the color that
it can detect (red, green, blue). The computer then looks at the
entire grid of sensors and their values, and calculates what the value
of each pixel (the squares in our screen-door example) should be. This
can get very complex right away, so I won't go into any deeper level
of detail than this; just remember that this is a form of
interpolation used just to figure out what color and brightness each
pixel should already have in the first place.
Also, you'll have seen that some digital cameras have optical zoom,
some have 'digital zoom', and some have a combination of the two.
Optical zoom is exactly that; the optical lenses in the camera are
used to zoom into the picture before the picture ever reaches the
sensors, so as far as the sensors are concerned, it's like you walked
closer to the picture itself. With digital zoom, though, it's all
interpolation. When you ask the camera to zoom in, it simply stretches
out a smaller portion of the 'big picture', and then makes educated
guesses to try to fill in the details between the original pixels.
Does this explanation help? If there is any part of this that is still
'fuzzy' to you, please let me know using the Request Clarification
button above!
Regards,
aht-ga
Google Answers Researcher |
