Thesis (Ph.D.)--City University of Hong Kong, 2009. / "Submitted to Department of Computer Science in partial fulfillment of the requirements for the degree of Doctor of Philosophy." Includes bibliographical references (leaves 145-161)

Thesis (M. Phil.)--Hong Kong University of Science and Technology, 2003. / Includes bibliographical references (leaves 107-113). Also available in electronic version. Access restricted to campus users.

Includes bibliographical references (p. 94-95).

Thesis--University of Alberta. / Title from screen (viewed Mar. 26, 2003). Includes bibliographical references.

Thesis (M. Phil.)--University of Hong Kong, 1989.

This dissertation describes a methodology for automated recognition of complex human activities. The dissertation presents a general framework which reliably recognizes various types of high-level human activities including human actions, human-human interactions, human-object interactions, and group activities. Our approach is a description-based approach, which enables a user to encode the structure of a high-level human activity as a formal representation. Recognition of human activities is done by semantically matching constructed representations with actual observations. The methodology uses a context-free grammar (CFG) based representation scheme as a formal syntax for representing composite activities. Our CFG-based representation enables us to define complex human activities based on simpler activities or movements. We have constructed a hierarchical framework which automatically matches activity representations with input observations. In the low-level of the system, image sequences are processed to extract poses and gestures. Based on the recognition of gestures, the high-level of the system hierarchically recognizes complex occurring human activities by searching for gestures that satisfies the temporal, spatial, and logical structure described in the representation. The concept of hallucinations and a probabilistic semantic-level recognition algorithm is introduced to cope with imperfect lower-layers. As a result, the system recognizes human activities including 'fighting', 'assault', 'a person leaving a suitcase', and 'a group of thieves stealing an object from owners', which are high-level activities that previous systems had difficulties. The experimental results show that our system reliably recognizes sequences of various types of complex human activities with a high recognition rate. / text

With the development and production of integrated circuits at the 22nm node, optical lithography faces increasing challenges to keep up with the specifications on its performance along various metrics, such as pattern fidelity and process window. The past few years saw the emergence of source mask optimization (SMO) as an important technique in computational lithography, which allows lithographers to rise to the challenges by exploiting a larger design space on both mask and illumination configuration, and integrates with methods such as inverse imaging. Yet, many methods that are used to tackle SMO problem arising in the inverse imaging involve heavy computation and slow convergence, making the technique unappealing for full-chip simulations or large circuits. Therefore, the purpose of this research is to take advantage of computational imaging technologies to solve source and mask design problems, extending the lifetime of optical lithography. The computational burden results in part from identical optimization over the whole mask pattern, consequently, we propose a weighted SMO scheme which applies different degrees of correction in the corresponding regions so that the optimal solutions are reached with fewer iterations. Additionally, undesirably long time is also attributed to the algorithm adopted to solve SMO problem. A fast algorithm based on augmented Lagrangian methods is therefore developed, which use the quasi-Newton method to accelerate convergence, thereby shortening the overall execution time. However, as semiconductor lithography is pushed to even smaller dimensions, mask topography effects have to be taken into account for a more accurate solution of SMO. At this stage, intensive computation is spent mainly in rigorous 3D mask modeling and simulations. To address this issue, we devise an optimization framework incorporating pupil aberrations into SMO procedure, which is performed based on the thin mask model so as to ensure a faster speed. We apply the above approaches to various mask geometries with different critical dimensions. Compared to conventional SMO, simulation results show that the proposed methods lead to better pattern fidelity and larger process window, especially in rigorous calculations. This demonstrates that the source and mask design generated through our algorithms are more practical. More importantly, the improved performance is not at the cost of speed. Instead, our methods take the least time to achieve it. This allows the advantages of computational imaging technologies to be worth exploring for further applications in optical lithography. / published_or_final_version / Electrical and Electronic Engineering / Doctoral / Doctor of Philosophy

Optical Coherence Tomography (OCT) is an optical tomography technique which provides high resolution non-invasive three-dimensional (3D) structural images of the sample based on coherent properties of light. The dissertation focuses on the use of OCT systems for detecting glaucoma, which is the second leading cause of blindness worldwide. First, as a prerequisite of analyzing ophthalmologic OCT images, a retinal sublayer segmentation algorithm is presented and implemented with GPU assisted computation. Then, a polarization-sensitive optical coherence tomography (PS-OCT) system was constructed for the study of glaucoma. Three closely related clinical and animal studies on early-stage glaucoma detection using either OCT or PS-OCT were performed. Statistical analysis of the study results indicates that the scattering property of retinal nerve fiber layer (RNFL) is the earliest indicator for glaucoma. Finally, to investigate the scattering properties of RNFL, a pathlength-multiplexed scattering-angle-diverse optical coherence tomography (PM-SAD-OCT) system was designed and built. PM-SAD-OCT images were collected from human and rodent retina as well as earthworm nerve cord. PM-SAD-OCT system shows promising potentials to detect neurodegenerative diseases including glaucoma. / text

published_or_final_version / Electrical Engineering / Master / Master of Philosophy

published_or_final_version / Electrical and Electronic Engineering / Doctoral / Doctor of Philosophy