Computers have many different techniques in image recognition.  The one I

am going to talk about image recognition through Edge Mapping.  Basically

the way it works is the computer captures an image, which is stored as values

for each pixel, or picture element.  The computer scans the set of values for

each pixel, and tries to determine any patterns using various formulas.  As you

can see in the image below, every picture contains a certain amount of noise.

So one method a computer could accommodate for this is to only register the

extremes.  These are the pixel values that over a certain limit, or group pixel

values into different groups.  Like this, different variations, or lack of

substance can easily get colored over and wiped away from the picture to

reduce some noise.  The reason this must be done is because unlike the human

eye, the computer cannot see through hazy images.  Humans can determine

objects even though there is a lot of noise in front of it, like a fog.  We know

through our own experiences what an object is even though there is something

distorting our view.  Mentally, we filter out the noise and fill in the blanks.

This is essentially what the computer is trying to do.  It reduces the noise, and

only focuses on the more predominant features, or the extremes.  These

mathematical formulas, such as “threshholding” or “quantizing”, reduce the

noise in the image such as the images below.  The first picture shows the actual

image captured by the computer.  The second picture shows the image after it

has been filtered.  By grouping pixel values, it could concentrate on the more

important features.  And finally the third picture shows the image after an

edge-mapping algorithm has been applied to it.  Basically, it shows the areas in

the image were there was major contrast in values.  This way it gives you an

outline of what the image actually looks like.  Now the computer has an image

too look at, study, and make comparisons with.