Alan Forghani
presented in Forghani (1997a). Based upon qualitative and quantitative assessment of edge detection results, a threshold
value of 40% was chosen to apply over the entire image of the study area which can be seen in Table 1. Morphological
transformation was undertaken over this data for generation of the classification map.
Title: USERS\ALI\PHD\FIGUR
Creator: JASC, Inc.
CreationDate: 5/6/1997
(a) 180 row by 160 column of a rural site
Title: USERS\ALI\PHD\URBAN\URBAN. eps
Creator: 'JASC, Inc.
CreationDate: 6/20/1997
(b) 185 row by 235 column of a built-up area
Figure 1. Test images; sub-sections of the 1982 image
3.2.3 Morphological Transformation over Edge Detection Data
The mathematical morphology tools are exploited after edge detection and thresholding to improve the edge detection
output. To clarify the functions of mathematical tools, the results of these operations are illustrated in Table 1. After
finding the best filters, threshold, the image analysis process (eg using Canny and MO) has been performed over the
entire image of the study area which can be seen in Figure 2a and 2b.
Five basic transformations of binary image are used in this process in the following order: bridge, fill, close, dilation,
and skeletonization. A single iteration was used for each operation. Bridge was used to bridge previously unconnected
pixels; fill isolated interior pixels; close performed binary closure in the data; dilation was used to add 8-connected
pixels to the boundary of binary objects. It helps to join edge segments within the binary image; the thinning eg
skeletonization) algorithm keeps the connectivity of the lines on an image. These operations increased the accuracy of
linear feature detection.
292 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.