Thresholding of Medical Images - Medical Image Processing: Advanced Fuzzy Set Theoretic Techniques - page 109

6

Thresholdin g of Medical Images

6.1 Introduction
Segmentation is a fundamental building block in image processing analysis.
It is the first stage of analysing an image. It partitions the image into disjoint
regions, which correlate strongly with objects or features of interest where
pixels of similar attributes are grouped together. Segmentation techniques
may be non-contextual or contextual. In the non-contextual technique, fea-
tures of an image are not taken into account and the pixels are grouped
together on the basis of a global attribute, that is, pixel intensities. It does
not take into account the relative location of the pixels. The contextual
technique takes into account the closeness of the pixels in an object. That
is, it exploits the relationship between the image features. Thresholding
is a simple and non-contextual technique. It is computationally inexpen-
sive and fast. Thresholding is a segmentation technique that classifies the
pixels into two categories: those pixels that fall below the threshold and
those that fall above the threshold. It involves analysing the histogram.
A threshold may be global or local. Global threshold selects a threshold
for the entire image. This method does not work if there is uneven illu-
mination, and in that case, local threshold works well. In local thresholds,
thresholds are obtained from each subregion of an image, thereby adapting
to local variations.
A complete segmentation of an image R is a finite set of regions R 1 , R 2 ,
R 3 , …, R N , such that

N

= ∪

RR RR i = and ∩ = φ , ≠ j

i i j

i

109