content-of the image, while the edges contribute to the high

frequencies. If the high frequencies are removed from the image

through spatial filtering, the sharp edges disappear, the large

continuous areas blend together smoothly, and the resultant image

appears soft or blurred. A low-pass image may not provide sufficient

resolution to discriminate between two similar objects. If the low

frequencies are removed from an image, the continuous areas become

dark with only the edges remaining. The image appears sharp with

well-defined edges and detail. This high-pass image provides, to the

human eye, the same or better discrimination of the original image.

That is, objects are identified and distinguished at least as well as

in the original image. For example, images containing a bright square

area and bright circular area are easily distinguished as a square and

circle. If the high frequencies are removed, both square and circle

appear as blobs with no distinct edges. However, if the low

frequencies are removed, the bright area in the center of the square

and circle disappears, leaving only a bright edge. Yet these bright

edges clearly indicate a square and a circle as shown in Figure 4.1.

Even if the square is not filled in, the edge clearly denotes the

square shape. The edge of the circular area still defines a circle.

The square and circle are easily distinguished in the high-pass

images. The information that distinguishes the square from the circle

is contained in the high frequencies.

 The traditional matched filter, as outlined in Chapter II, is

created from the complex conjugate of the Fourier transform of the

reference image. Filtering with such a filter is equivalent to

correlating the reference image with a test image. Because most