Geometric Model Reconstruction - Computational Hemodynamics: Theory, Modelling and Applications - page 48

Threshold level = 20

original CT-scanned image

Threshold level = 128 Threshold level = 175

Fig. 3.4 A simple threshold criterion applied to the scanned coronary artery image. Greyscale
values of the pixels range from 0 to 255 where 0 is typically set as black and 255 is set as white

summarise threshold algorithms are suggested, (Glasbey 1993; Le et al. 1990; Sa-
hoo et al. 1988; Sezgin and Sankur 2004).

3.3.3 Edge Based Segmentation

Threshold algorithms are highly dependent on the defined criteria. A narrow or
stringent criterion may lead to loss of pixels that exists in the region of interest,
while conversely a loose criterion may lead to inclusion of non-required regions.
A scanned object, such as the artery passageway exhibits a change in its structure
between the inner blood flow, the lumen wall, and the surrounding tissue, and this
is reflected through the greyscale levels on the image. If the boundaries of an indi-
vidual structure can be detected, then the enclosed region containing the structure,
can be separated and segmented from the scan. This boundary detection is a form of
edge based segmentation techniques.
Edge-based segmentation represents a large group of methods based on the idea
that an edge is defined as an area in an image where the intensity changes rapidly.
This is achieved by applying a filter over the image to detect the rapid change in
pixel values and thereby labelling the pixel as part of an edge or non-edge. Seg-
mentation is performed by allocating a single label category of all non-edge pixels
which are contiguous. Edge detection of an image significantly reduces the amount