we attempt to locate directly the boundaries that exist between the regions or
we seek to identify edge pixels and then link them together to establish the
required boundaries. Segmentations resulting from the two approaches may not
be exactly the same, but both approaches are useful to understanding and
solving image segmentation problems, and their combined use can lead to
improved performance [2–4]. Real-world applications in digital microscopy
often pose very challenging segmentation problems. Variations and combi-
nations of the basic techniques presented here often must be tailored to the
specific application to produce acceptable results.
9.1.1 Pixel Connectivity
Before introducing various methods for image segmentation, it is important to
understand the concept of connectivity of pixels in a digital image (see also
Chapters 8 and 14). In a set of connected pixels, all the pixels are adjacent or
touching [5]. Between any two pixels in a connected set there exists a connected
path wholly within the set. A connected path is one that always moves between
neighboring pixels. Thus, in a connected set, one can trace a connected path
between any two pixels without ever leaving the set.
There are two rules of connectivity. If only laterally adjacent pixels (up, down,
right, left) are considered to be connected, we have ‘‘4-connectivity,’ and the
objects are ‘‘4-connected.’’ Thus a pixel has only four neighbors to which it can
be connected. If diagonally adjacent (458 neighbor) pixels are also considered to be
connected, then we have ‘‘8-connectivity,’ the objects are ‘‘8-connected,’’ and each
pixel has eight neighbors to which it can be connected. Either connectivity rule can
be adopted as long as it is used consistently. Any region that is 4-connected is also
8-connected, but the converse is not necessarily true. Overall, 8-connectivity is
more commonly used, and it produces results that are closer to one’s intuition.
9.2 Region-Based Segmentation
Region segmentation methods partition an image by grouping similar pixels to-
gether into identified regions. Image content within a region should be uniform and
homogeneous with respect to certain attributes, such as intensity, rate of change in
intensity, color, and texture. Regions are important in interpreting an image
because they typically correspond to objects or parts of objects in a scene. In this
section we discuss a number of widely used techniques that fall into this category.
9.2.1 Thresholding
Thresholding is an essential region-based image segmentation technique that is
particularly useful for scenes containing solid objects resting on a contrasting
9 Image Segmentation
160

Get Microscope Image Processing now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.