for datasets With noh-sp

Segmentati
Using Clustering for Image
4.5
Segmentation
4.5.1 What is Image an image into
involves dividing
fundamental task in image processing that The goal of
Image segmentation is a representation and make it easier to analyze.
to simplity its correspond toobjects or
areas
multiple segments or regions meaningrul parts that
partitionan image into
image segmentation is to
of interest within the image. boundaries (lines
important intormation, identify objects,
we can extract The result of image
By segmenting an image, and enable further analysis and processing tasks.
Curves, etc.), and
textures,
segments
that collectively cover the entire image, or a set of boundaries
Segmentation is a set of
Each of the pixels
in a region is similar with respect to some characteristic,
extracted from the image. texture.
intensity,or
Such as color. Use Cases
Segmentation
Image
Use Case
Images
Detection in MRI
1. Tumor
imaging,
MRI Scans are used to visualize internal structures of the body, including detecting
In medical
segmentation plays a vital role in identifying and delineating tumors from surrounding
Image
tumors.

the MRI scan, doctors can accurately locate and analyze the size,
in MRIimages. By segmenting
characteristics of the tumor for diagnosis and treatment planning.
tissues

shape,
and
 Machine Learning
4.20
in the field of autonomous vehicles
is essential for tasks
By segmenting different
segmentation

2.
Autonomous
Vehicles

detection, lane
: Image
detection, and
obstacle
avoidance.

mnarkings, and
traffic signs,. autonomous vehicles
elements
in the
like object pedestrians, road is often used to classify can make
scene such as
vehicles,
navigation.
Semantic
segmentation

understand
its each pixel in
informed
decisions for safe
categories,
the vehicle to
enabling surroundings and react
predefined
an image into driving.
accordingly toensure safe
Segmnentation
Works?
4.5.2 How Image into distinct regions or segments, each
Image
segmentation involves
dividing an image
segmentation enables
the represented
selective processing of specific
by masks or labeled
images. This
than rather processingthe entire image. image
for targeted analysis
segments. It is useful segmentation is to
detect abrupt changes in
technique in image pixel values,
" which
One often correspondtoedges that delineate different regions within the image. These edges
common

serve as boundaries between areas with varying characteristics, aiding in the segmentation
procesS.
identifying similarities among regions in an image. Techniques
" Another approach involves utilized to group pixels with siml
such as region growing, clustering, and thresholding are
based on shared characteristics
attributes together, forming coherent segments
been developed to tackle segmentotie.
Over time, a variety of domain-specific approaches have
leverage domain
challenges in specific application domains effectively. These tailored methods
knowledge to address the unique requirements and characteristics of different types of
images and
segmnentation.
segmentation tasks. K-Means clustering is a method commonly used for image

4.5.3 K-Means Clustering for Image Segmentation

K-Means clustering is a popular unsupervised machine learning algorithm used for clustering data
points into a specified number of clusters. In the context of image segmentation, K-Means clustering
similarity. By
can be applied to group pixels in an image into distinct clusters based on their feature
clustering pixels based on features like color intensity, texture, or spatial proximity, we can identify
of
regions in an image that share common properties. This helps in separating objects or areas
interest from the background.

How K-Means Clustering Works for Image Segmentation

partition an
The K-means algorithm is a popular clustering technique used in image segmentation to
image into K clusters based on pixel similarities.
Algorithm K-Means Clustering for Image Segmentation
centers in the feature space,
1. Initialization: The algorithm starts by randomly initializing Kcluster
where Kis the predefined number of clusters. in
centroid is closestto it
2. Assignment Step: Each pixel in theimage is assigned to the cluster whose used to measureesimilarity.
terms of feature similarity. The distance metric, often Euclidean distance,, is based recalculated

clusters are
3. Update Step: After assigning all pixels to clusters, the centroids ofthe
on the mean of the pixel values within each cluster.
 repeated iteratively until convergence criteria are met, such as a
2 and 3 are maximum
Iteration:
Steps

or
mal
minim a l change in cluster assignments.
iterations

L of
Segmentation: Oncethe
number
algorithm converges,
each pixel in the image is associated with the cluster it
effectively segmented the image into distinct regions based on pixel similarity.
5. It is
to.
helongs