image produces an obvious improvement in quality to a human observer.

Image restoration techniques compensate an image, which has been

degraded in some fashion, to restore it as nearly as possible to its

undegraded state. For example, an image which is blurred due to

camera motion may be improved using motion restoration. To perform

the difficult task of image interpretation, extraneous noise must be

separated from the desired signals. This may occur in several stages

of enhancement where each stage reduces the extraneous noise and

preserves the information crucial to object recognition. Image

enhancement may include contrast transformation, frame subtraction,

and spatial filtering. The goal of image enhancement is to reduce the

image complexity so that feature analysis is simplified.1

 Once the scene has been enhanced, the job of interpretation is

simplified. The interpreter must now decide what the remaining

features represent. The features present a pattern to the interpreter

to be recognized. This pattern recognition problem may be quite

difficult when a large number of features are necessary to

differentiate between two possibilities. Most people have to look

closely to see any difference between two twins. A computer might

have equal difficulty distinguishing a car from a house in low-

resolution image.

 Recognition involves an interpretation of an image. This includes

scene matching and understanding. Scene matching determines which

region in an image is similar to a pictorial description of a region

of another scene. A reference region or template is provided and

systematically compared to each region in a larger image. Here the

computer attempts to match models of known objects, such as cars,