Efficient graph-based image segmentation for natural images
Other Title(s)
تقسيم الصور الطبيعية الفعال باستخدام التمثيل البياني
Dissertant
Thesis advisor
Comitee Members
University
Middle East University
Faculty
Faculty of Information Technology
Department
Computer Science Department
University Country
Jordan
Degree
Master
Degree Date
2017
English Abstract
Image segmentation is the process of partitioning an input image into multiple segments (sets of pixels, also known as super-pixels).
The widespread utilization of segmentation in various real-time applications has urged the need to reduce the segmentation cost without compromising the efficiency of the produced output.
Graphbased is one of the effective techniques for image segmentation.
Although, graph-based uses a greedy approach to illuminate connected edges between distinctive regions in several iterations, it produces satisfactory results compares to the existing segmentation approaches.
However, graph-based technique consumes time, which make it inapplicable for real-time applications.
This thesis proposed an extended graph-based image segmentation technique by reduce the segmentation time of the original technique and improve the accuracy of the output results.
To achieve the aforementioned goal, the process, which is implemented over a constructed graph corresponding to an input image, forms homogenous regions around initial points that are selected automatically.
The initial points, which characterized as significant spots in the image that lies within a coherent region, are determined based on the intensity of the neighborhood pixels.
By using initial points, the number of iterations that is required to complete the segmentation process is decreased, which will decrease the running time.
Moreover, based on the initial points, the regions that are produced in the segmentation process are coherent since it is accumulated over coherent pixels, which will enhance the accuracy of the output results.
The experiments were conducted on Pascal-challenge images.
The results showed that the proposed segmentation technique gain an accuracy of 84%, which is higher than the original approach that achieved an accuracy of 68%.
Moreover, the proposed segmentation approach reduces the running time of the segmentation by 62% compared to the original approach.
Main Subjects
Information Technology and Computer Science
Topics
No. of Pages
76
Table of Contents
Table of contents.
Abstract.
Abstract in Arabic.
Chapter One : Introduction.
Chapter Two : Literature review and related work.
Chapter Three : The proposed methodology.
Chapter Four : Experimental results and analysis.
Chapter Five : Conclusion and future work.
References.
American Psychological Association (APA)
al-Mughir, Layla Khalil. (2017). Efficient graph-based image segmentation for natural images. (Master's theses Theses and Dissertations Master). Middle East University, Jordan
https://search.emarefa.net/detail/BIM-762683
Modern Language Association (MLA)
al-Mughir, Layla Khalil. Efficient graph-based image segmentation for natural images. (Master's theses Theses and Dissertations Master). Middle East University. (2017).
https://search.emarefa.net/detail/BIM-762683
American Medical Association (AMA)
al-Mughir, Layla Khalil. (2017). Efficient graph-based image segmentation for natural images. (Master's theses Theses and Dissertations Master). Middle East University, Jordan
https://search.emarefa.net/detail/BIM-762683
Language
English
Data Type
Arab Theses
Record ID
BIM-762683