dc.contributor.author |
Baydoun, Mohammed Hussein, |
dc.date |
2014 |
dc.date.accessioned |
2015-02-03T10:23:59Z |
dc.date.available |
2015-02-03T10:23:59Z |
dc.date.issued |
2014 |
dc.date.submitted |
2014 |
dc.identifier.other |
b18292033 |
dc.identifier.uri |
http://hdl.handle.net/10938/10052 |
dc.description |
Dissertation. Ph.D. American University of Beirut. Department of Electrical and Computer Engineering, 2014. ED:53 |
dc.description |
Advisor : Dr. Mohammed Adnan Al-Alaoui, Professor, Electrical and Computer Engineering ; Chairman of Committee : Dr. Nesreene Ghaddar, Professor, Mechanical Engineering ; Members of Committee: Dr. Ali Chehab, Associate Professor, Electrical and Computer Engineering ; Dr. Fadi Karameh, Associate Professor, Electrical and Computer Engineering ; Dr. Daniel Asmar, Associate Professor, Mechanical Engineering ; Dr. Lina Karam, Professor, Electrical and Computer Engineering, Arizona State University, USA ; Dr. Haidar Harmanani, Professor, Computer Science, Lebanese American University, Lebanon ; Dr. Maha El Choubassi, Doctor, Electrical and Computer Engineering, Intel, USA ; Dr. Rony Ferzli, Doctor, Electrical and Computer Engineering, Intel, USA. |
dc.description |
Includes bibliographical references (leaves 154-167) |
dc.description.abstract |
This work aims at providing an accurate and fast stereo matching system. Stereo matching targets determining the depths of the pixels in an image(s) obtained through a stereo setup. Stereo matching is one of the most addressed problems of image processing. The purpose here is achieved using varying notions that might seem disconnected, but they will be developed and related throughout the whole work. The different ideas mainly depend on using different image processing techniques and classification. At first, the stereo images are treated using histogram information to reduce color differences. Afterwards, initial stereo matching is considered using various approaches with emphasis on basic ones that utilize measures such as the Sum of Absolute Distances (SAD) and Normalized Cross Correlation (NCC). Furthermore, edge based segmentation is proposed to improve on stereo matching, where the edge detection uses digital differentiator approximation. Besides, different classification approaches are used to enhance stereo matching using a set of proposed features that are depicted from the stereo pair and the preliminary stereo matching. The classification methods, not exclusive to others, include boosting and others. And in order to advance boosting and similar methods, some modifications are proposed based on training and selecting classifiers. These propositions are experimentally proven to generally enhance the accuracy and performance of classification whether regarding stereo matching or otherwise. The stereo images are primarily selected from the Middlebury stereo database to show the validity of the various propositions in accordance with previous literature. A basic time analysis is provided for many of the ideas especially that speed is a major factor in stereo matching. Compute Unified Device Architecture (CUDA) on Nvidia Graphics Processing Units (GPUs) is used to show the real time performance of several of the suggested notions. Moreover, several possibilities are discussed to provide directions tow |
dc.format.extent |
1 online resource (xv, 167 leaves) : illustrations (some color) ; 30cm |
dc.language.iso |
eng |
dc.relation.ispartof |
Theses, Dissertations, and Projects |
dc.subject.classification |
ED:000053 AUBNO |
dc.subject.lcsh |
Image processing. |
dc.subject.lcsh |
Computer vision. |
dc.subject.lcsh |
Graphics processing units. |
dc.subject.lcsh |
Computer algorithms. |
dc.subject.lcsh |
CUDA (Computer architecture) |
dc.subject.lcsh |
Artificial intelligence. |
dc.subject.lcsh |
Signal processing. |
dc.subject.lcsh |
Parallel programming (Computer science) |
dc.title |
Enhanced stereo matching using improved classification - |
dc.type |
Dissertation |
dc.contributor.department |
American University of Beirut. Faculty of Engineering and Architecture. Department of Electrical and Computer Engineering, degree granting institution. |