dc.contributor.author |
Jalloul, Manal Khalil |
dc.date.accessioned |
2017-08-30T14:12:40Z |
dc.date.available |
2017-08-30T14:12:40Z |
dc.date.issued |
2016 |
dc.date.submitted |
2015 |
dc.identifier.other |
b18692722 |
dc.identifier.uri |
http://hdl.handle.net/10938/10852 |
dc.description |
Dissertation. Ph.D. American University of Beirut. Department of Electrical and Computer Engineering, 2016. ED:71 |
dc.description |
Chair of Committee : Dr. Ayman Kayssi, Professor, Electrical and Computer Engineering ; Advisor : Dr. Mohamad Adnan Al-Alaoui, Professor, Electrical and Computer Engineering ; Members of Committee : Dr. Ali Chehab, Professor, Electrical and Computer Engineering ; Dr. Haitham Akkary, Professor, Electrical and Computer Engineering ; Dr. Daniel Asmar, Associate Professor, Mechanical Engineering ; Dr. Ali Sayed, Professor, Electrical Engineering, University of California, Los Angeles, USA ; Dr. Haidar Harmanani, Professor, Computer Science, Lebanese American University, Lebanon ; Dr. Maha El Choubassi, Senior Research Scientist, Intel Corporation, USA ; Dr. Rony Ferzli, Soft Architect, Intel Corporation, USA. |
dc.description |
Includes bibliographical references (leaves 161-169) |
dc.description.abstract |
Motion estimation (ME) is a common tool used in all video coding standards. Fast and accurate algorithms are needed to target the real-time processing requirements of emerging applications. On the other hand, in the hardware industry, there is great emphasis on High Performance Computing (HPC) which is characterized by a shift to multi and many core systems. The programming community has to embrace the new parallelism in order to take advantage of the performance gains offered by the new technology. The block motion estimation (BME) problem is classified as non-convex since the objective function is multimodal. Existing fast block matching methods suffer from poor accuracy and are susceptible to being trapped into local optima on the error surface. The collective intelligence enabled by the particle swarm optimization (PSO) technique, however, was found effective in alleviating the local optima problem. Belonging to the category of evolutionary algorithms, PSO is capable of handling non-differentiable, discontinuous and multimodal objective functions. To this end, in this dissertation, several efficient and parallel ME algorithms based on PSO are proposed. Several levels of parallelisms are introduced into the ME process. First, parallelism between the macroblocks (MBs) of the frame is achieved through a novel cooperative ME scheme based on a multi-swarm PSO model that performs ME in a cooperative manner concurrently for all the MBs in the frame. Several strategies are incorporated into the dynamics of the PSO algorithm to improve its motion estimation accuracy and enhance its convergence speed including a novel initialization scheme, a fitness function history preservation algorithm, and a dynamically varied maximum velocity. The multi-core and GPU implementations of the proposed framework showed that the speedup provided is scalable with the video resolution. Second, parallelism is introduced within the MB through two different approaches based on distributed multi-agents systems. The problem of BME is first c |
dc.format.extent |
1 online resource (xviii, 169 leaves) : illustrations |
dc.language.iso |
eng |
dc.relation.ispartof |
Theses, Dissertations, and Projects |
dc.subject.classification |
ED:000071 |
dc.subject.lcsh |
Evolutionary computation. |
dc.subject.lcsh |
Video compression. |
dc.subject.lcsh |
Parallel algorithms. |
dc.subject.lcsh |
CUDA (Computer architecture) |
dc.subject.lcsh |
Multiagent systems. |
dc.subject.lcsh |
Game theory. |
dc.subject.lcsh |
Evolutionary programming (Computer science) |
dc.title |
On parallel, distributed, and hybrid evolutionary algorithms for block motion estimation - |
dc.type |
Dissertation |
dc.contributor.department |
Faculty of Engineering and Architecture. |
dc.contributor.department |
Department of Electrical and Computer Engineering, |
dc.contributor.institution |
American University of Beirut. |