Image segmentation based on normalised cuts with clustering algorithm

Choong, Mei Yeen (2013) Image segmentation based on normalised cuts with clustering algorithm. Masters thesis, Universiti Malaysia Sabah.

	Text 24 PAGES.pdf Download (311kB)
	Text FULLTEXT.pdf Restricted to Registered users only Download (9MB)

Abstract

This research presents an image segmentation approach based on normalised cuts. Normalised cuts based image segmentation is well known for its reliability to produce good segmentation. It does segmentation on an image by representing the image as weighted graph in global view. The normalised cuts method plays a role in partitioning the graphs into sub-graphs. Clustering algorithm is then implemented to the sub-graphs’ corresponding pixels that contain common similarities are grouped together into one cluster and then dissimilar pixels are separated into distinctive clusters. However, to build the weighted graph, similarity measurement of all possible paired pixels is complicated. Thus, an approach to facilitate the normalised cuts algorithm is performed for effective image segmentation to reduce the complexity of the similarity measurement. This can be done by performing the normalised cuts algorithm in hierarchical manner in which the image is divided into image cells where local segmentation is performed on each of the image cells. Clustered segments from the local segmentation are then used for image segmentation in global perspective. As the clusters initialisation gives influence to the segmentation result, optimisation of the clustering algorithm is implemented to achieve a better segmentation. With the approach applied in the normalised cuts based image segmentation, the constraint of using normalised cuts algorithm in image segmentation can be alleviated. Evaluation of c -means and fuzzy c-means clustering algorithm with normalised cuts image segmentation on various kinds of images has been carried out. Results show that fuzzy c-means clustering has an improvement of 14.96% over c -means clustering in obtaining accurate segmentation.

Item Type:	Thesis (Masters)
Keyword:	Image segmentation, Normalized cuts, Clustering algorithms, K-means, Fuzzy k-means
Subjects:	T Technology > TA Engineering (General). Civil engineering (General) > TA1-2040 Engineering (General). Civil engineering (General) > TA1501-1820 Applied optics. Photonics
Department:	SCHOOL > School of Engineering and Information Technology
Depositing User:	DG MASNIAH AHMAD -
Date Deposited:	15 Apr 2025 10:53
Last Modified:	15 Apr 2025 10:53
URI:	https://eprints.ums.edu.my/id/eprint/43486

Actions (login required)

View Item