Image Segmentation: A Rough-set Theoretic Approach - Rough Fuzzy Image Analysis: Foundations and Methodologies - page 189

10

Image Segmentation: A Rough-set
Theoretic Approach

10.1 Introduction :::::::::::::::::::::::::::::::::::::::::::: 10{1
10.2 Rough-set theory and properties :::::::::::::::::::: 10{2
10.3 The Concept of Histon ::::::::::::::::::::::::::::::: 10{3
Construction of Histon Roughness measure
10.4 Segmentation Method :::::::::::::::::::::::::::::::: 10{6
Selection of peak and Threshold values Region
Merging
10.5 Experimental Results ::::::::::::::::::::::::::::::::: 10{7
10.6 Summary ::::::::::::::::::::::::::::::::::::::::::::::: 10{10
Bibliography ::::::::::::::::::::::::::::::::::::::::::::::::: 10{14

Milind M. Mushrif
YeshwantraoChavanCollegeofEngineering,
Nagpur

Ajoy K. Ray
IndianInstituteofTechnology,Kharagpur

10.1Introduction

Image segmentation is a critical, yet essential task in many applications. Segmentation
subdivides an image into a set of homogeneous and meaningful regions, such that the pixels
in each partitioned region possess identical set of properties or attributes and the union
of any two adjacent regions is non-homogeneous (Gonzalez and Woods, 2002). The set of
properties may include gray levels, color, contrast, spectral values, or textural properties.
The color image segmentation has gained paramount importance in recent times largely
due to the availability inexpensive digital cameras, increasing computational power of the
computers and decreasing cost of computation. The applications of color image segmen-
tation include medical image diagnostics, video object segmentation, object based video
compression, object detection from remotely sensed images, and many more.
The number of dierent approaches exist in the literature for the color image segmen-
tation. They can be broadly classied into histogram based, edge based, region based,
clustering, graph theoretic, rule-based or knowledge driven and combination of these tech-
niques (Aghbari and Al-Haj, 2006; Cheng and Li, 2003; Acharya and Ray, 2005). Though
there are a large number of segmentation algorithms available, there is no algorithm that
can be considered to be good for all the images (Pal and Pal, 1993). One of the most widely
used technique for image segmentation is the histogram based thresholding, which assumes
that homogeneous objects in the image manifest themselves as clusters. These methods do
not need any a-priori information of the image (Liu and Yang, 1994), but they do not take
into account the spatial correlation of the same or similar valued elements. However, The
real-world images usually have strong correlation among the neighboring pixels. The fuzzy

10{1