Enhanced Clustering Algorithms For Gray-Scale Image Segmentation

Simple item page
dc.contributor.authorSiddiqui, Fasahat Ullah
dc.date.accessioned2018-07-11T01:08:13Z
dc.date.available2018-07-11T01:08:13Z
dc.date.issued2012-04
dc.description.abstractThe clustering algorithms are widely used as an unsupervised method for image segmentation in medical diagnosis, satellite imaging and biometric systems. The algorithms are chosen since they are easy to be implemented, required low computational time and less sensitive to noise and artifacts. However, in some cases the conventional clustering algorithms introduce over-segmentation problems and unable to preserve the region of interest (i.e. objects). The first problem occurs because of certain clustering algorithms are sensitive to outliers, meanwhile the latter problem is due to the dead centroids and trapped centroids at non-active regions phenomenon. Thus, four new clustering algorithms namely the Optimized K-Means (OKM), Enhanced Moving K-Means-1(EMKM-1), Enhanced Moving K-Means-2(EMKM-2) and Outlier Rejection Fuzzy C-Means (ORFCM) are proposed. The OKM algorithm has a capability to differentiate between the empty and the zero variance clusters. The assignment of the data to these clusters are redesigned from that of the conventional K-Means clustering algorithm. On the other hand, the EMKM-1 and EMKM-2 algorithms reformed the data transferring concept of the conventional Adaptive Moving K-Means (AMKM) which is able to avoid the aforementioned problems. Furthermore, the ORFCM algorithm employed the adaptable exponent operator of Euclidean distance in fuzzy membership function in order to improve the capability of the Fuzzy C-Means (FCM) by minimizing the overlapping regions to moderate the outlier effects. The performance of the proposed clustering algorithms is compared qualitatively and quantitatively with several state- of-the-art clustering algorithms. The results conclude that the proposed clustering algorithms outperform the conventional clustering algorithms by producing more homogenous regions in an image with better in shape and sharp edges preservation. In addition, the OKM algorithm attains the best results among the other tested algorithms.en_US
dc.identifier.urihttp://hdl.handle.net/123456789/5896
dc.language.isoenen_US
dc.publisherUniversiti Sains Malaysiaen_US
dc.subjectThe clustering algorithms are widely used asen_US
dc.subjectan unsupervised method for image segmentationen_US
dc.titleEnhanced Clustering Algorithms For Gray-Scale Image Segmentationen_US
dc.typeThesisen_US
Files
License bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
license.txt
Size:
1.71 KB
Format:
Item-specific license agreed upon to submission
Description:
Download
Collections
Pusat Pengajian Kejuruteraaan Elektrik dan Elektronik - Tesis