ABSTRACT

Figure 1.5 shows additional segmentations using this algorithm on several of the images in Figure 1.1. It should be noted that when multispectral data is available (e.g., RGB) the algorithm can be applied to each band. Figure 1.5(a,b) shows the result of applying our algorithm to the

a

c

b

Extracting Straight Lines and Line-Based Texture Features

a

c

e

f

g

Thresholding Simple thresholding algorithms typically partition an image by assign-

32 HANSON AND RISEMAN

The rule-based region merging algorithm starts with an initial segmentation as input and selects pairs of adjacent regions, which are candidates for merging into a single region. For example, the initial segmentation can be produced by a region growing algorithm with an extremely conservative threshold that produces a highly fragmented segmentation. Because the system only merges (as opposed to splitting) regions, the initial segmentation must contain all the region boundaries desired in the final result.