Page 265 - Digital Analysis of Remotely Sensed Imagery

Image Enhancement 227

–1 2 1 –1 2 1
–1 2 1 –2 4 2
–1 2 1 –1 2 1
(a) (b)
FIGURE 6.17 The Laplacian templates based on second-order derivatives.
(a) Unweighted line; (b) weighted line.

FIGURE 6.18 Output image that has been Laplacian edge-enhanced using
the raw image in Fig. 6.13a. Conspicuous in the image are linear features
that are oriented in a certain direction.
vertically, or diagonally to detect edges oriented in a direction per-
pendicular to them. As with all edge-detection templates, the nine
coefficients are designed so that they add up to zero. This sum of
coefficients should not be used in the convolution shown in Eq. (6.8).
The Laplacian template is another edge-detection-operant based
on second-order derivatives. They are effective at detecting lines or
spots distinctively from ramp edges (Leica, 2006). They have two
forms, one used to detect unweighted lines (Fig. 6.17a) and one used
to detect weighted lines (Fig. 6.17b). The processed images often show
edges and zeros conspicuously (Fig. 6.18). In this enhanced image,
large pixel values are much larger, and small pixel values are
much smaller than those in the input image.

6.5 Multiple-Image Manipulation
The aforementioned enhancement techniques involve only a single
band in the input and in the output. In practice it is possible to gener-
ate a new image from multiple images of the same area. Usually,
these images are multispectral bands from the same sensor obtained
at the same time, covering an identical geographic area. If any of
these conditions is not met (e.g., images from different sensors or
obtained from the same sensor at different times), the images need to