Development and performance evaluation of a distributed image processing model running on LAN-based systems

العناوين الأخرى

تطوير و تقييم أداء نظام توزيعي لمعالجة الصور باستخدام شبكات الحاسوب المحلية

مقدم أطروحة جامعية

Abu Hammad, Maysun Yusuf

مشرف أطروحة جامعية

al-Bahadili, Husayn

أعضاء اللجنة

al-Ani, Muzhir Shaban
Hassan, Shakir

الجامعة

جامعة عمان العربية

الكلية

كلية العلوم الحاسوبية و المعلوماتية

القسم الأكاديمي

قسم علم الحاسوب

دولة الجامعة

الأردن

الدرجة العلمية

ماجستير

تاريخ الدرجة العلمية

2008

الملخص الإنجليزي

Image processing applications involve different processes, such as: image enhancement, edge detection, object detection, noise removal, color quantization, etc.

Image processing applications are characterized as time and memory demanding applications, and, for many practical and strategic applications, needs to be speeded up by using faster computing systems. The main objective of this work is to develop and evaluate the performance of a Distributed Image Processing (DIP) model that can be used to perform extensive image processing computations, on continuous feed-in images, on a LAN-Based computer system.

The model is based on the widely-used processor-farm parallel methodology, in which each processor executes the same program independently from other processors, each operating on a different part of the total data.

Therefore, no interprocessor communication is required other than that involved in forwarding data/results to and from the processors. For equivalent calculations, the computation times on a single processor (PC) and a LAN-based computing system of various number of processors (PCs), are compared, and the resulting speedup (S), parallelization efficiency (E), and image processing rate () are quoted.

For the LAN-based system, the effect of varying the number of PCs, the number of images and execution time algorithm on the overall system speedup, efficiency, and image processing rate, are estimated. The LAN-based DIP model demonstrates an excellent performance for recourse-demanding image processing applications in terms of simplicity, adaptability, flexibility, expandability, transparency, and high efficiency.

It efficiently utilizes the computational power of a LAN-Based system using a Java Parallel Virtual Machine (JPVM) library as a data communication toolset for message passing between PCs, where, for standard image processing computations, it achieves parallelization efficiencies that vary between 87% to 60% when the number of PCs varies between 2 to 5 PCs connected through 10/100 Mbps Ethernet switch.

التخصصات الرئيسية

الهندسة الكهربائية

الموضوعات

• الإلكترونيات الرقمية

عدد الصفحات

66

قائمة المحتويات

• APA
• MLA
• AMA

لغة النص

الإنجليزية

نوع البيانات

رسائل جامعية

رقم السجل

BIM-529036