No description available.

Chan Ka Cheong. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2004. / Includes bibliographical references (leaves 77-85). / Abstracts in English and Chinese. / Chapter 1. --- Introduction --- p.1 / Chapter 1.1 --- Introduction to Fingerprint Verification --- p.1 / Chapter 1.1.1 --- Biometrics --- p.1 / Chapter 1.1.2 --- Fingerprint History --- p.2 / Chapter 1.1.3 --- Fingerprint characteristics --- p.4 / Chapter 1.1.4 --- A Generic Fingerprint Matching System Architecture --- p.6 / Chapter 1.1.5 --- Fingerprint Verification and Identification --- p.8 / Chapter 1.1.7 --- Biometric metrics --- p.10 / Chapter 1.2 --- Embedded system --- p.12 / Chapter 1.2.1 --- Introduction to embedded systems --- p.12 / Chapter 1.2.2 --- Embedded systems characteristics --- p.12 / Chapter 1.2.3 --- Performance evaluation of a StrongARM processor --- p.13 / Chapter 1.3 --- Objective -An embedded fingerprint verification system --- p.16 / Chapter 1.4 --- Organization of the Thesis --- p.17 / Chapter 2 --- Literature Reviews --- p.18 / Chapter 2.1 --- Fingerprint matching overviews --- p.18 / Chapter 2.1.1 --- Minutiae-based fingerprint matching --- p.20 / Chapter 2.2 --- Fingerprint image enhancement --- p.21 / Chapter 2.3 --- Orientation field Computation --- p.22 / Chapter 2.4 --- Fingerprint Segmentation --- p.24 / Chapter 2.5 --- Singularity Detection --- p.25 / Chapter 2.6 --- Fingerprint Classification --- p.27 / Chapter 2.7 --- Minutia extraction --- p.30 / Chapter 2.7.1 --- Binarization and thinning --- p.30 / Chapter 2.7.2 --- Direct gray scale approach --- p.32 / Chapter 2.7.3 --- Comparison of the minutiae extraction approaches --- p.35 / Chapter 2.8 --- Minutiae matching --- p.37 / Chapter 2.8.1 --- Point matching --- p.37 / Chapter 2.8.2 --- Structural matching technique --- p.38 / Chapter 2.9 --- Summary --- p.40 / Chapter 3. --- Implementation --- p.41 / Chapter 3.1 --- Fast Fingerprint Matching System Overview --- p.41 / Chapter 3.1.1 --- Typical Fingerprint Matching System --- p.41 / Chapter 3.1.2. --- Fast Fingerprint Matching System Overview --- p.41 / Chapter 3.2 --- Orientation computation --- p.43 / Chapter 3.21 --- Orientation computation --- p.43 / Chapter 3.22 --- Smooth orientation field --- p.43 / Chapter 3.3 --- Fingerprint image segmentation --- p.45 / Chapter 3.4 --- Reference Point Extraction --- p.46 / Chapter 3.5 --- A Classification Scheme --- p.51 / Chapter 3.6 --- Finding A Small Fingerprint Matching Area --- p.54 / Chapter 3.7 --- Fingerprint Matching --- p.57 / Chapter 3.8 --- Minutiae extraction --- p.59 / Chapter 3.8.1 --- Ridge tracing --- p.59 / Chapter 3.8.2 --- cross sectioning --- p.60 / Chapter 3.8.3 --- local maximum determination --- p.61 / Chapter 3.8.4 --- Ridge tracing marking --- p.62 / Chapter 3.8.5 --- Ridge tracing stop criteria --- p.63 / Chapter 3.9 --- Optimization technique --- p.65 / Chapter 3.10 --- Summary --- p.66 / Chapter 4. --- Experimental results --- p.67 / Chapter 4.1 --- Experimental setup --- p.67 / Chapter 4.2 --- Fingerprint database --- p.67 / Chapter 4.3 --- Reference point accuracy --- p.67 / Chapter 4.4 --- Variable number of matching minutiae results --- p.68 / Chapter 4.5 --- Contribution of the verification prototype --- p.72 / Chapter 5. --- Conclusion and Future Research --- p.74 / Chapter 5.1 --- Conclusion --- p.74 / Chapter 5.2 --- Future Research --- p.74 / Bibliography --- p.77

Lau Tak Fu. / Thesis submitted in: October 2003. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2004. / Includes bibliographical references (leaves 93-98). / Abstracts in English and Chinese. / Chapter 1 - --- Introduction --- p.9 / Chapter 1.1 --- Scope of the work --- p.9 / Chapter 1.2 --- Main contribution --- p.11 / Chapter 1.3 --- Outline of the thesis --- p.12 / Chapter 2 - --- Background --- p.14 / Chapter 2.1 --- Three dimensional models from images --- p.14 / Chapter 2.2 --- Un-calibrated 3D reconstruction --- p.14 / Chapter 2.3 --- Self calibrated 3D reconstruction --- p.16 / Chapter 2.4 --- Initial model formation using image based --- p.18 / Chapter 2.5 --- Volumes from Silhouettes --- p.19 / Chapter 3 - --- Initial model reconstruct the problem with mismatch noise --- p.22 / Chapter 3.1 --- Perspective Camera Model --- p.24 / Chapter 3.2 --- "Intrinsic parameters, Extrinsic parameters and camera motion" --- p.25 / Chapter 3.2.1 --- Intrinsic parameters --- p.25 / Chapter 3.2.2 --- Extrinsic parameter and camera motion --- p.27 / Chapter 3.3 --- Lowe's method --- p.29 / Chapter 3.4 --- Interleave bundle adjustment for structure and motion recovery from multiple images --- p.32 / Chapter 3.5 --- Feature points mismatch analysis --- p.38 / Chapter 4 - --- Feature selection by using look forward silhouette clipping --- p.43 / Chapter 4.1 --- Introduction to silhouette clipping --- p.43 / Chapter 4.2 --- Silhouette clipping for 3D model --- p.45 / Chapter 4.3 --- Implementation --- p.52 / Chapter 4.3.1 --- Silhouette extraction program --- p.52 / Chapter 4.3.2 --- Feature filter for alternative bundle adjustment algorithm --- p.59 / Chapter 5 - --- Experimental data --- p.61 / Chapter 5.1 --- Simulation --- p.61 / Chapter 5.1.1 --- Input of simulation --- p.61 / Chapter 5.1.2 --- Output of the simulation --- p.66 / Chapter 5.1.2.1 --- Radius distribution --- p.66 / Chapter 5.1.2.2 --- 3D model output --- p.74 / Chapter 5.1.2.3 --- VRML plotting --- p.80 / Chapter 5.2 --- Real Image testing --- p.82 / Chapter 5.2.1 --- Toy house on a turntable test --- p.82 / Chapter 5.2.2 --- Other tests on turntable --- p.86 / Chapter 6 - --- Conclusion and discussion --- p.89

Wang, Guangyu. / Thesis submitted in: December 2003. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2004. / Includes bibliographical references (leaves 70-75). / Abstracts in English and Chinese. / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- BDCT Compression Artifacts --- p.1 / Chapter 1.2 --- Previous Artifact Removal Methods --- p.3 / Chapter 1.3 --- Our Method --- p.4 / Chapter 1.4 --- Structure of the Thesis --- p.4 / Chapter 2 --- Related Work --- p.6 / Chapter 2.1 --- Image Compression --- p.6 / Chapter 2.2 --- A Typical BDCT Compression: Baseline JPEG --- p.7 / Chapter 2.3 --- Existing Artifact Removal Methods --- p.10 / Chapter 2.3.1 --- Post-Filtering --- p.10 / Chapter 2.3.2 --- Projection onto Convex Sets --- p.12 / Chapter 2.3.3 --- Learning by Examples --- p.13 / Chapter 2.4 --- Other Related Work --- p.14 / Chapter 3 --- Contamination as Markov Random Field --- p.17 / Chapter 3.1 --- Markov Random Field --- p.17 / Chapter 3.2 --- Contamination as MRF --- p.18 / Chapter 4 --- Training Set Preparation --- p.22 / Chapter 4.1 --- Training Images Selection --- p.22 / Chapter 4.2 --- Bit Rate --- p.23 / Chapter 5 --- Artifact Vectors --- p.26 / Chapter 5.1 --- Formation of Artifact Vectors --- p.26 / Chapter 5.2 --- Luminance Remapping --- p.29 / Chapter 5.3 --- Dominant Implication --- p.29 / Chapter 6 --- Tree-Structured Vector Quantization --- p.32 / Chapter 6.1 --- Background --- p.32 / Chapter 6.1.1 --- Vector Quantization --- p.32 / Chapter 6.1.2 --- Tree-Structured Vector Quantization --- p.33 / Chapter 6.1.3 --- K-Means Clustering --- p.34 / Chapter 6.2 --- TSVQ in Artifact Removal --- p.35 / Chapter 7 --- Synthesis --- p.39 / Chapter 7.1 --- Color Processing --- p.39 / Chapter 7.2 --- Artifact Removal --- p.40 / Chapter 7.3 --- Selective Rejection of Synthesized Values --- p.42 / Chapter 8 --- Experimental Results --- p.48 / Chapter 8.1 --- Image Quality Assessments --- p.48 / Chapter 8.1.1 --- Peak Signal-Noise Ratio --- p.48 / Chapter 8.1.2 --- Mean Structural SIMilarity --- p.49 / Chapter 8.2 --- Performance --- p.50 / Chapter 8.3 --- How Size of Training Set Affects the Performance --- p.52 / Chapter 8.4 --- How Bit Rates Affect the Performance --- p.54 / Chapter 8.5 --- Comparisons --- p.56 / Chapter 9 --- Conclusion --- p.61 / Chapter A --- Color Transformation --- p.63 / Chapter B --- Image Quality --- p.64 / Chapter B.1 --- Image Quality vs. Quantization Table --- p.64 / Chapter B.2 --- Image Quality vs. Bit Rate --- p.66 / Chapter C --- Arti User's Manual --- p.68 / Bibliography --- p.70

Ip Che Yin. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2004. / Includes bibliographical references (leaves 119-127). / Abstracts in English and Chinese. / Abstract --- p.i / Acknowledgement --- p.iii / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- Motivation --- p.1 / Chapter 1.2 --- Previous Work --- p.2 / Chapter 1.2.1 --- Reconstruction for Architecture Scene --- p.2 / Chapter 1.2.2 --- Super-resolution --- p.4 / Chapter 1.2.3 --- Coarse-to-Fine Approach --- p.4 / Chapter 1.3 --- Proposed solution --- p.6 / Chapter 1.4 --- Contribution --- p.6 / Chapter 1.5 --- Publications --- p.7 / Chapter 1.6 --- Layout of the thesis --- p.7 / Chapter 2 --- Background Techniques --- p.8 / Chapter 2.1 --- Interest Point Detectors --- p.8 / Chapter 2.1.1 --- Scale-space --- p.9 / Chapter 2.1.2 --- Harris Corner detectors --- p.10 / Chapter 2.1.3 --- Other Kinds of Interest Point Detectors --- p.17 / Chapter 2.1.4 --- Summary --- p.18 / Chapter 2.2 --- Steerable filters --- p.19 / Chapter 2.2.1 --- Orientation estimation --- p.20 / Chapter 2.3 --- Point Descriptors --- p.22 / Chapter 2.3.1 --- Image derivatives under illumination change --- p.23 / Chapter 2.3.2 --- Image derivatives under geometric scale change --- p.24 / Chapter 2.3.3 --- An example of a point descriptor --- p.25 / Chapter 2.3.4 --- Other examples --- p.25 / Chapter 2.4 --- Feature Tracking Techniques --- p.26 / Chapter 2.4.1 --- Kanade-Lucas-Tomasi (KLT) Tracker --- p.26 / Chapter 2.4.2 --- Guided Tracking Algorithm --- p.28 / Chapter 2.5 --- RANSAC --- p.29 / Chapter 2.6 --- Structure-from-motion (SFM) Algorithm --- p.31 / Chapter 2.6.1 --- Factorization methods --- p.33 / Chapter 2.6.2 --- Epipolar Geometry --- p.39 / Chapter 2.6.3 --- Bundle Adjustment --- p.47 / Chapter 2.6.4 --- Summary --- p.50 / Chapter 3 --- Hierarchical Registration of 3D Models --- p.52 / Chapter 3.1 --- Overview --- p.53 / Chapter 3.1.1 --- The Arrangement of Image Sequences --- p.53 / Chapter 3.1.2 --- The Framework --- p.54 / Chapter 3.2 3 --- D Model Reconstruction for Each Sequence --- p.57 / Chapter 3.3 --- Multi-scale Image Matching --- p.59 / Chapter 3.3.1 --- Scale-space interest point detection --- p.61 / Chapter 3.3.2 --- Point descriptor --- p.61 / Chapter 3.3.3 --- Point-to-point matching --- p.63 / Chapter 3.3.4 --- Image transformation estimation --- p.64 / Chapter 3.3.5 --- Multi-level image matching --- p.66 / Chapter 3.4 --- Linkage Establishment --- p.68 / Chapter 3.5 --- 3D Model Registration --- p.70 / Chapter 3.6 --- VRML Modelling --- p.73 / Chapter 4 --- Experiment --- p.74 / Chapter 4.1 --- Synthetic Experiments --- p.74 / Chapter 4.1.1 --- Study on Rematching Algorithm --- p.74 / Chapter 4.1.2 --- Comparison between Affine and Metric transforma- tions for 3D Registration --- p.80 / Chapter 4.2 --- Real Scene Experiments --- p.86 / Chapter 5 --- Conclusion --- p.112 / Chapter 5.1 --- Future Work --- p.114 / Chapter A --- Camera Parameters --- p.116 / Chapter A.1 --- Intrinsic Parameters --- p.116 / Chapter A.2 --- Extrinsic Parameters --- p.117 / Bibliography --- p.127

by Yuan, Ding. / Thesis submitted in: December 2003. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2004. / Includes bibliographical references (leaves 69-71). / Abstracts in English and Chinese. / ABSTRACT --- p.i / 摘要 --- p.iii / ACKNOWLEDGEMENTS --- p.v / TABLE OF CONTENTS --- p.vi / LIST OF FIGURES --- p.viii / LIST OF TABLES --- p.xii / Chapter 1 --- INTRODUCTION --- p.1 / Chapter 2 --- PREVIOUS WORK --- p.5 / Chapter 2.1 --- Traditional Stereo Vision --- p.5 / Chapter 2.1.1 --- Epipolar Constraint --- p.7 / Chapter 2.1.2 --- Some Constraints Based on Properties of Scene Objects --- p.9 / Chapter 2.1.3 --- Two Classes of Algorithms for Correspondence Establishment --- p.10 / Chapter 2.2 --- Correspondenceless Stereo Vision Algorithm for Single Planar Surface Recovery under Parallel-axis Stereo Geometry --- p.13 / Chapter 3 --- CORRESPONDENCE-FREE STEREO VISION UNDER GENERAL STEREO SETUP --- p.19 / Chapter 3.1 --- Correspondence-free Stereo Vision Algorithm for Single Planar Surface Recovery under General Stereo Geometry --- p.20 / Chapter 3.1.1 --- Algorithm in Its Basic Form --- p.21 / Chapter 3.1.2 --- Algorithm Combined with Epipolar Constraint --- p.25 / Chapter 3.1.3 --- Algorithm Combined with SVD And Robust Estimation --- p.36 / Chapter 3.2 --- Correspondence-free Stereo Vision Algorithm for Multiple Planar Surface Recovery --- p.45 / Chapter 3.2.1 --- Plane Hypothesis --- p.46 / Chapter 3.2.2 --- Plane Confirmation And 3D Reconstruction --- p.48 / Chapter 3.2.3 --- Experimental Results --- p.50 / Chapter 3.3 --- Experimental Results on Correspondence-free Vs. Correspondence Based Methods --- p.60 / Chapter 4 --- CONCLUSION AND FUTURE WORK --- p.65 / APPENDIX --- p.67 / BIBLIOGRAPHY --- p.69

Xu, Yi. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2011. / Includes bibliographical references (p. 62-69). / Abstracts in English and Chinese. / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- Organization --- p.4 / Chapter 2 --- Background and Motivation --- p.7 / Chapter 2.1 --- ID Mondrian Smoothing --- p.9 / Chapter 2.2 --- 2D Formulation --- p.13 / Chapter 3 --- Solver --- p.16 / Chapter 3.1 --- More Analysis --- p.20 / Chapter 4 --- Edge Extraction --- p.26 / Chapter 4.1 --- Related work --- p.26 / Chapter 4.2 --- Method and Results --- p.28 / Chapter 4.3 --- Summary --- p.32 / Chapter 5 --- Image Abstraction and Pencil Sketching --- p.35 / Chapter 5.1 --- Related Work --- p.35 / Chapter 5.2 --- Method and Results --- p.36 / Chapter 5.3 --- Summary --- p.40 / Chapter 6 --- Clip-Art Compression Artifact Removal --- p.41 / Chapter 6.1 --- Related work --- p.41 / Chapter 6.2 --- Method and Results --- p.43 / Chapter 6.3 --- Summary --- p.46 / Chapter 7 --- Layer-Based Contrast Manipulation --- p.49 / Chapter 7.1 --- Related Work --- p.49 / Chapter 7.2 --- Method and Results --- p.50 / Chapter 7.2.1 --- Edge Adjustment --- p.51 / Chapter 7.2.2 --- Detail Magnification --- p.54 / Chapter 7.2.3 --- Tone Mapping --- p.55 / Chapter 7.3 --- Summary --- p.56 / Chapter 8 --- Conclusion and Discussion --- p.59 / Bibliography --- p.61

A real-time dynamic feedback control system is developed. An improved PID algorithm, called the integral separated piecewise PID scheme, is used in the control system. This algorithm is able to limit the contribution of the integral component in the PID calculation to avoid integral windup. In addition, it could use different PID parameters to adapt to different segments within one punch motion cycle. Hence, the error of the punch motion, either resulting from the machine assembly or from the machine dynamics, can be compensated by tuning the velocity of the servomotor. This is a unique feature of the new press that ensures its accuracy. / Based on the novel structure, the detailed design is then carried out, which includes the mechanical design, kinematics and inverse kinematics analysis, static force analysis, parametric design and the other related designs. A calibration method based on the experiment and computer simulation is proposed for the new press, which is also useful for the parallel mechanisms. Cooperated with Guangdong Metal Forming Machine Works Co. Ltd., a 250 KN prototype has been built and tested. / In order to ensure the desirable performance, dynamic control is necessary. The thesis uses two dynamic modeling methods to study the dynamics of the press. One is the kineto-static method. It is also called D'Alembert principle which rearranges Newton's second law and transfers a dynamic problem to an equivalent static problem by adding the inertial forces and inertial torques onto the system. The model can then be analyzed easily and exactly as a static system subjected to the inertial forces and torques and the external forces. The other method is the Lagrangian method which derives the dynamic model from the energy perspective. Based on the model, the dynamics of the press is studied by means of computer simulation and is validated experimentally. / In this thesis, a controllable hybrid mechanical metal forming press is developed, which is driven by a CSM with a flywheel and a servomotor. From a mechanism point of view, it is a closed-loop 2-DOF parallel planar five-bar mechanism with four resolute joints and one prismatic joint. Thanks to the usage of the servomotor, the punch motion of the new press can be controlled by tuning the velocity of the servomotor. Accordingly, desired punch motions for different stamping processes can be obtained. In other words, the new press is flexible and controllable like the servo mechanical press and the hydraulic press. Moreover, the CSM with flywheel provides the main power during the stamping operation, and hence, it is energy efficient. In addition, it is not expensive to build, as it uses only a small servomotor. / Metal forming is one of the oldest production processes and yet, is still one of the most commonly used processes today. Everyday, millions of parts are produced by metal forming ranging from battery caps to automobile body panels. Therefore, even a small improvement may add to significant corporative gain. / The thesis also describes the trajectory planning method for the press, which is based on the combination of the inverse kinematics and cubic spline interpolation. The trajectory is optimized under multiple constraints on velocity, acceleration and jerk of the servomotor. It guarantees the new press is controllable and energy efficient. / Two typical stamping processes, drawing and forging, are taken as examples for the operations of the new press. The results of the simulation and the experiment match well. Based on the simulation and experiments, it is found that the CSM provides the main power for the metal forming operations, while the servomotor is mainly responsible for overcoming the inertia forces to realize the desired punch motion. The experiments show that the new press is energy efficient, fast, controllable and inexpensive to build. It combines the advantages of both mechanical press and hydraulic press and has a good performance. It is expected the new press will have a great potential for the metal forming industry. (Abstract shortened by UMI.) / He, Kai. / "February 2008." / Adviser: Ruxu Du. / Source: Dissertation Abstracts International, Volume: 70-03, Section: B, page: 1902. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2008. / Includes bibliographical references (p. 147-149). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Electronic reproduction. [Ann Arbor, MI] : ProQuest Information and Learning, [200-] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstracts in English and Chinese. / School code: 1307.

In Part I of the thesis, we study the problems of image denoising and de-blurring under impulse noise. We consider two-phase methods for solving these problems. In the first phase, efficient detectors are applied to detect the outliers. In the second phase, variational methods utilizing the outputs of the first phase are performed. For denoising, we prove that the functionals to be minimized in the second phase have many good properties such as maximum principle, Lipschitz continuity and etc. Based on the results, we propose conjugate gradient methods and quasi-Newton methods to minimize the functional efficiently. For deblurring, we propose a two-phase method combining the median-type filters and a variational method with Mumford-Shah regularization term. The experiments show that the two-phase methods give much better results than both the median-type filters and full variational methods. / In this thesis, we study two aspects in image processing. Part I is about image denoising and deblurring under impulse noise, and Part II is about framelet-based methods for image reconstruction. / Part II of the thesis focuses on framelet-based methods for image reconstruction. In particular, we consider framelet-based methods for chopped and nodded image reconstruction and image inpainting. By interpreting both the problems as recovery of missing data, framelet, a generalization of wavelet, is applied to solve the problems. We incorporate sophisticated thresholding schemes into the algorithm, hence the regularities of the restored images can be guaranteed. By the theory of convex analysis, we prove the convergence of the framelet-based methods. We find that the limits of the framelet-based methods satisfy some minimization properties, hence connections with variational methods are established. / Cai, Jianfeng. / "March 2007." / Adviser: Raymond H. Chan. / Source: Dissertation Abstracts International, Volume: 69-01, Section: B, page: 0350. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2007. / Includes bibliographical references (p. 119-129). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Electronic reproduction. [Ann Arbor, MI] : ProQuest Information and Learning, [200-] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstracts in English and Chinese. / School code: 1307.

Lastly, in the context of motion deblurring, we discuss a few new motion deblurring problems that are significant to blur kernel estimation and nonblind deconvolution. We found that strong edges do not always profit kernel estimation, but instead under certain circumstance degrade it. This finding leads to a new metric to measure the usefulness of image edges in motion deblurring and a gradient selection process to mitigate their possible adverse effect. It makes possible to solve for very large blur PSFs which easily fail existing blind deblurring methods. We also propose an efficient and high-quality kernel estimation method based on the spatial prior and the iterative support detection (ISD) kernel refinement, which avoids hard threshold of the kernel elements to enforce sparsity. We employ the TV-ℓ1 deconvolution model, solved with a new variable substitution scheme to robustly suppress noise. / This thesis covers complete discussion of motion estimation and deblurring and presents new methods to tackle them. In the context of motion estimation, we study the problem of estimating 2D apparent motion from two or more input images, referred to as optical flow estimation. We discuss several specific fundamental problems in existing optical flow estimation frameworks, including 1) estimating flow vectors for textureless and occluded regions, which was regarded as infeasible and with large ambiguities, and 2) the incapability of the commonly employed coarse-to-fine multi-scale scheme to preserve motion structures in several challenging circumstances. / To address the problem of multi-scale estimation, we extend the coarse-to-fine scheme by complementing the initialization at each scale with sparse feature matching, based on the observation that fine motion structures, especially those with significant and abrupt displacement transition, cannot always be correctly reconstructed owing to an incorrect initialization. We also introduce the adaption of the objective function and development of a new optimization procedure, which constitute a unified system for both large- and small-displacement optical flow estimation. The effectiveness of our method is borne out by extensive experiments on small-displacement benchmark dataset as well as the challenging large-displacement optical flow data. / To further increase the sub-pixel accuracy, we study how resolution changes affect the flow estimates. We show that by simple upsampling, we can effectively reduce errors for sub-pixel correspondence. In addition, we identify the regularization bias problem and explore its relationship to the image resolution. We propose a general fine-to-coarse framework to compute sub-pixel color matching for different computer vision problems. Various experiments were performed on motion estimation and stereo matching data. We are able to reduce errors by up to 30%, which would otherwise be very difficult to achieve through other conventional optimization methods. / We propose novel methods to solve these problems. Firstly, we introduce a segmentation based variational model to regularize flow estimates for textureless and occluded regions. Parametric and Non-parametric optical flow models are combined, using a confidence map to measure the rigidity of the moving regions. The resulted flow field is with high quality even at motion discontinuity and textureless regions and is very useful for applications such as video editing. / Xu, Li. / Adviser: Jiaya Jia. / Source: Dissertation Abstracts International, Volume: 73-03, Section: B, page: . / Thesis (Ph.D.)--Chinese University of Hong Kong, 2010. / Includes bibliographical references (leaves 126-137). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Electronic reproduction. [Ann Arbor, MI] : ProQuest Information and Learning, [201-] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstract also in Chinese.