論文研究了圖像及技頻信號的去躁技術，共分為四部分。第一部分研究了用來改進編碼的圖像的主觀同客觀賞質量的方法，介紹了兩個有效的圖像后處理技術。第二部分研究了基於貝葉斯估計鋒的對高斯噪聲圖像的去噪方法。第三部分介紹了對編碼的視頻信號去噪的兩個方法，這些方法可作為視頻環路和后處理濾波器，從而改進視頻信號的編碼性能。第四部分介紹了圖像邊緣檢測技術，該技術可結合基於於邊緣的圖像處理設術來改進含噪圖像的主觀同客觀質量。 / 第一部分，我們研究了一些用來改進編碼的圖像和視頻主觀同客觀質量的有效技術。採用基於塊的離散餘弦變換(DCT) 壓縮標準的圖像和視頻存在編碼效應。為了減少圖像的編碼效應，我們提出了一個基於綫性最小均值平方誤差(MMSE)標準的非局部Kuan (NLK) 濾波器。它被用來恢復量化后的DCT係數，然後我們提出了雙重量非局部Kuan (DNLK) 濾波器。這個DNLK濾波器用在過完備的DCTl係數，進一步形成了過完備雙重非局部Kuan (OCNLK) 濾波器。通過對一些使用測試用的量化表編碼的圖像， JPEG編碼的圖像的實驗，驗證了這兩种方法的有效性。 / 第二部分，我們採用了貝葉斯估計器來估計原始的變換域係數。該方法是對非局部Kuan (NLK) 濾波器的改進，它採用了高斯尺度混合模型來考慮變換域係數的非高斯特性。除去非局部估計的均值后的殘差係數並不是高斯分佈，所以我們採用了高斯混合模型來表示表示這些殘差係數。實驗給果證明了它在圖像去噪應用上的有效性，以及比NLK 濾波器更好的性能。 / 第三部分描述了用來改進視頻編碼性能的基於NLK 的方法。對於採用楨内/間預測和變換編碼的視頻而言，直接使用NLK濾波器將不能有效的去除編碼效應。我們分析了其中的原因，並提出了兩個視頻環路濾波器:基於四叉樹的NLK (QNLK) 濾波器和基於四叉樹的過完備NLK (QOCNLK) 濾波器。NLK 和過完備NLK濾波器用來恢復視頻殘差的量化係數。這些恢復后的係數進一步投射到所設計的量化約數集(QCS)裏。四叉樹技術用來進行自適應的濾波控制。實驗結果表明了相對于H.264/AVC HighProfile ，這些技術達到了可觀的比特率壓縮效率和視覺質量改善。我們同時也採用了QNLK和QOCNLK濾波器進行視頻后處理，並給出了實驗結果和分析。 / 第四部分，我們研究了一個基於非下採樣小波，使用三維隱馬爾可伕模型(HMM)的邊緣檢測技術。 提出的模型不僅能捕獲不同尺度間小波係數的關係，而且考慮了尺度内係數的依賴性。一個計算有效的最大似然(ML) 估計算法用來計算該模型的參數，每一個係數對應的隱含狀態用最大后騐(MAP) 估計來求得。基於一些自然圖像的實驗給果用來評償該算法。對於含噪圖像，該算法可以同時提取邊緣和去除噪聲信號。另外，提出的該模型對於其它的圖像視頻處理任務也是一個潛在的多尺度統計模型工具。 / The thesis investigates techniques to reduce noise in image and video signals.The investigation can be divided into four parts. The first part studies methods to improve the objective and subjective quality of coded images. The proposed two efficient image postprocessing techniques are presented. The second part investigates noise reduction of image signals corrupted by Gaussian noise and presents a denoising method based on Bayes estimator. The third part investigates noise reduction of coded videos by two different approaches. The approaches are used as video loop and postprocessing filters, in order to improve the coding performance of video signals. Finally, the fourth part investigates a new edge detection method that can provide an accurate detection of edges. The method can be used together with edge-based image processing techniques to improve objective and subjective quality of noisy images. / In the first part, the technique to improve the subjective and objective quality of coded images are studied. Coding artifacts exist in images using block-based discrete cosine transform (DCT) compression standards. In order to reduce image coding artifacts, a non-local Kuan's (NLK) filter is proposed from the minimum mean-square-error (MMSE) criterion. It is used to restore quantized DCT coefficients. Then we propose the dual non-local Kuan's (DNLK) filter by applying the NLK filter in dual-layer. The DNLK filter is further extended to form the overcomplete dual non-local Kuan's (OCDNLK) filter by applying to the overcomplete DCT coefficients. Experimental results on coded images using test quantization tables and JPEG coded images show the effectiveness of the two methods. / In the second part, we use a Bayes least square estimator to estimate the original transform coefficients optimally, from the Bayes perspective. It is an improved nonlocal Kuan's filter by considering non-Gaussian property of transform coefficients using Gaussian scale mixture model. The residual coefficients after subtracting non-local estimated means are not Gaussian distributed, so a Gaussian scale mixture model is employed to represent the residual coefficients. Experiments demonstrate its efficiency on image denoising and better performance than the NLK filter. / In the third part, the proposed methods based on the NLK filter to improve video coding performance are studied. Direct application of the NLK filter on videos coded using intra/inter-frame prediction and transform coding cannot improve coding performance efficiently. We identify the causes of the problem and propose quadtree-based NLK (QNLK) loop filter and quadtree-based overcomplete NLK (QOCNLK) loop filter to solve the problem. NLK and overcomplete NLK loop filters are used to restore quantized residual transform coefficients. Restored coefficients are then projected onto designed quantization constraint sets (QCS). Quadtree-based signaling strategy is used for adaptive filtering control. Experimental results show that the proposed loop filtering techniques achieve significant bit rate saving and visual quality improvement compared with H.264 or advanced video coding (AVC) High Profile. We also provide the experimental results and analysis of postprocessing application using QNLK and QOCNLK filters. / In the fourth part, a new edge detecting technique using 3-D Hidden Markov Model (HMM) based on the non-decimated wavelet is studied. The proposed model can not only capture the relationship of wavelet coefficients inter-scale, but also consider the intra-scale dependence. A computationally efficient maximum likelihood (ML) estimation algorithm is employed to compute parameters and the hidden state of each coefficient is revealed by maximum a posteriori (MAP) estimation. Experimental results of natural images are provided to evaluate the algorithm. For noisy images, the method can extract edges and remove noise simultaneously. The method can be used together with edge-based image processing techniques to improve subjective and objective quality of noisy images. In addition, the proposed model has the potential to be an efficient multi-scale statistical modeling tool for other image or video processing tasks. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Zhang, Renqi. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2012. / Includes bibliographical references (leaves 109-113). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstract also in Chinese. / Dedication --- p.ii / Acknowledgments --- p.iii / Abstract --- p.viii / Publications --- p.x / Nomenclature --- p.xii / Contents --- p.xviii / List of Figures --- p.xxi / List o f Tables --- p.xxiii / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- Motivation and Objectives --- p.1 / Chapter 1.2 --- Thesis Outline --- p.4 / Chapter 2 --- Image/Video Coding Standard --- p.5 / Chapter 2.1 --- Image Coding Standard JPEG --- p.5 / Chapter 2.1.1 --- JPEG Codec --- p.6 / Chapter 2.1.2 --- Block-based Discrete Cosine Transform (DCT) --- p.7 / Chapter 2.1.3 --- Quant ization and Entropy Coding --- p.9 / Chapter 2.1.4 --- Decoding --- p.11 / Chapter 2.1.5 --- Effects of JPEG Compression --- p.12 / Chapter 2.2 --- Video Coding Standard H.264/AVC --- p.12 / Chapter 2.2.1 --- H.264 Codec --- p.14 / Chapter 2.2.2 --- Inter Prediction --- p.15 / Chapter 2.2.3 --- Intra Prediction --- p.15 / Chapter 2.2.4 --- Transform and Quantization --- p.17 / Chapter 2.2.5 --- Deblocking Filter --- p.19 / Chapter 2.3 --- Conclusion --- p.21 / Chapter 3 --- Image Postprocessing --- p.22 / Chapter 3.1 --- Introduction --- p.22 / Chapter 3.2 --- The Proposed Non-local KUAN's Filter in the DCT Domain --- p.24 / Chapter 3.2.1 --- LMMSE Filter in the DCT Domain --- p.25 / Chapter 3.2.2 --- The solution of the LMMSE Filter --- p.26 / Chapter 3.2.3 --- Non-local Parameter Estimation --- p.29 / Chapter 3.2.4 --- Investigation on Assumptions 1 and 2 --- p.29 / Chapter 3.3 --- Image Postprocessing by the Dual Non-local Kuan's Filter --- p.34 / Chapter 3.3.1 --- DNLK Filter --- p.35 / Chapter 3.3.2 --- OCDNLK Filter --- p.36 / Chapter 3.4 --- Experimental Results and Analysis --- p.37 / Chapter 3.4.1 --- Experiments on Three Test Quantization Tables --- p.37 / Chapter 3.4.2 --- Experiment al Setting --- p.38 / Chapter 3.4.3 --- Experiment al Results and Comparison with State-of-the-Art Methods --- p.38 / Chapter 3.4.4 --- Postprocessing on JPEG Coded Images --- p.40 / Chapter 3.4.5 --- Complexity Analysis --- p.40 / Chapter 3.5 --- Conclusion --- p.41 / Chapter 4 --- Improved Non-local Kuan's Filter using a Gaussian Scale Mixture --- p.49 / Chapter 4.1 --- Introduction --- p.49 / Chapter 4.1 --- Non-local Kuan's filter in the DCT domain --- p.50 / Chapter 4.1 --- Improved Non-local Kuan's Filter using a Guassian Scale Mixture --- p.52 / Chapter 4.3.1 --- Gaussian Scale Mixture Model --- p.52 / Chapter 4.3.2 --- Computation of Bayes Least Square Estimator --- p.55 / Chapter 4.4 --- Experimental Sett ings and Results --- p.56 / Chapter 4.5 --- Conclusion --- p.57 / Chapter 5 --- Quadtree-based Non-local Kuan's Filtering in Video Compression --- p.59 / Chapter 5.1 --- Introduction --- p.59 / Chapter 5.2 --- The Proposed Loop Filters --- p.62 / Chapter 5.2.1 --- Non-local Kuan's Filter --- p.62 / Chapter 5.2.2 --- The NLK Loop Filter --- p.65 / Chapter 5.2.3 --- The Overcomplete NLK Loop Filter --- p.67 / Chapter 5.2.4 --- Use of t he Quant ization Constraint Set (QCS) --- p.69 / Chapter 5.2.5 --- Performance of t he NLK and OCNLK Loop Filters for Video Coding --- p.70 / Chapter 5.2.6 --- The Proposed Quadtree-based NLK and OCNLK Loop Filters --- p.74 / Chapter 5.3 --- Experimental Result s and Analysis of QNLK and QOCNLK Loop Filters --- p.76 / Chapter 5.3.1 --- Experiment al Results of QNLK and QOCNLK Loop Filters --- p.76 / Chapter 5.3.2 --- Computational Complexity Analysis --- p.78 / Chapter 5.4 --- QNLK and QOCNLK Postprocessing Filters --- p.82 / Chapter 5.4.1 --- Experimental Result s and Analysis ofQNLK and QOCNLK Postprocessing Filters --- p.84 / Chapter 5.5 --- Conclusion --- p.85 / Chapter 6 --- Image Edge Detection using 3-D Hidden Markov Model based on the Non-decimated Wavelet --- p.88 / Chapter 6.1 --- Introduction --- p.88 / Chapter 6.2 --- Three-dimensional HMM based on the Non-decimated Wavelet (3-D N-WHMM) --- p.90 / Chapter 6.2.1 --- Hidden Markov Model (HMM) --- p.90 / Chapter 6.2.2 --- 3-D Hidden Markov Model (3-D HMM) --- p.91 / Chapter 6.2.3 --- Construction of the 3-D NWHMM --- p.92 / Chapter 6.2.4 --- Probabilistic Model for Individual Wavelet Coefficient --- p.92 / Chapter 6.2.5 --- Training Paramet ers and Searching for Hidden States --- p.94 / Chapter 6.3 --- Experimental Results and Analysis --- p.96 / Chapter 6.3.1 --- The algorithm implementation --- p.96 / Chapter 6.3.2 --- Comparisons with Canny and two multi-scalet echniques --- p.96 / Chapter 6.3.3 --- Experimental results of noisy images --- p.98 / Chapter 6.4 --- Conclusion --- p.98 / Chapter 7 --- Conclusions and Future Work --- p.103 / Chapter 7.1 --- Contributions of The Thesis --- p.103 / Chapter 7.2 --- Future Work and Research Directions --- p.104 / Chapter A --- Appendix A --- p.107 / Chapter A.1 --- Mathematical Derivations of equation (4.6) --- p.107 / Bibliography --- p.109

Video is currently at the forefront of most business and natural environments. In surveillance, it is the most important technology as surveillance systems reveal information and patterns for solving many security problems including crime prevention. This research investigates technologies that currently drive video surveillance systems with a view to optimization and automated decision support. The investigation reveals some features and properties that can be optimised to improve performance and derive further benefits from surveillance systems. These aspects include system-wide architecture, meta-data generation, meta-data persistence, object identification, object tagging, object tracking, search and querying sub-systems. The current less-than-optimum performance is attributable to many factors, which include massive volume, variety, and velocity (the speed at which streaming video transmit to storage) of video data in surveillance systems. Research contributions are 2-fold. First, we propose a system-wide architecture for designing and implementing surveillance systems, based on the authors' system architecture for generating meta-data. Secondly, we design a simulation model of a multi-view surveillance system from which the researchers generate simulated video streams in large volumes. From each video sequence in the model, the authors extract meta-data and apply a novel algorithm for predicting the location of identifiable objects across a well-connected camera cluster. This research provide evidence that independent surveillance systems (for example, security cameras) can be unified across a geographical location such as a smart city, where each network is administratively owned and managed independently. Our investigation involved 2 experiments - first, the implementation of a web-based solution where we developed a directory service for managing, cataloguing, and persisting metadata generated by the surveillance networks. The second experiment focused on the set up, configuration and the architecture of the surveillance system. These experiments involved the investigation and demonstration of 3 loosely coupled service-oriented APIs – these services provided the capability to generate the query-able metadata. The results of our investigations provided answers to our research questions - the main question being “to what degree of accuracy can we predict the location of an object in a connected surveillance network”. Our experiment also provided evidence in support of our hypothesis – “it is feasible to ‘explore' unified surveillance data generated from independent surveillance networks”.

Lu, Shi. / Thesis submitted in: December 2004. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2005. / Includes bibliographical references (leaves 100-109). / Abstracts in English and Chinese. / Abstract --- p.ii / Acknowledgement --- p.vi / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- Motivation and Objectives --- p.1 / Chapter 1.2 --- Our Contributions --- p.3 / Chapter 1.3 --- Thesis Outline --- p.4 / Chapter 2 --- Related Work --- p.6 / Chapter 2.1 --- Static Video Summary --- p.6 / Chapter 2.2 --- Dynamic Video Skimming --- p.10 / Chapter 2.3 --- Summary --- p.14 / Chapter 3 --- Greedy Method Based Skim Generation --- p.16 / Chapter 3.1 --- Selected Video Features for Video Summarization --- p.17 / Chapter 3.2 --- Video Summarization Problem --- p.18 / Chapter 3.3 --- Experiments --- p.22 / Chapter 3.4 --- Summary --- p.25 / Chapter 4 --- Video Structure Analysis --- p.27 / Chapter 4.1 --- Video Shot Detection --- p.29 / Chapter 4.1.1 --- Shot Cut Detection --- p.30 / Chapter 4.1.2 --- Fade Detection --- p.35 / Chapter 4.2 --- Video Shot Group Construction --- p.38 / Chapter 4.2.1 --- Shot Pairwise Similarity Measure --- p.39 / Chapter 4.2.2 --- Video Shot Grouping by VToC --- p.41 / Chapter 4.2.3 --- Spectral Graph Partitioning --- p.42 / Chapter 4.3 --- Video Scene Detection --- p.46 / Chapter 4.4 --- Shot Arrangement Patterns --- p.48 / Chapter 4.5 --- Experiments --- p.50 / Chapter 4.6 --- Summary --- p.53 / Chapter 5 --- Graph Optimization-Based Video Summary Generation --- p.55 / Chapter 5.1 --- Video Scene Analysis --- p.56 / Chapter 5.1.1 --- Scene Content Entropy --- p.57 / Chapter 5.1.2 --- Target Skim Length Assignment --- p.58 / Chapter 5.2 --- Graph Modelling of Video Scenes --- p.59 / Chapter 5.2.1 --- Decompose the Video Scene into Candidate Video Strings --- p.60 / Chapter 5.2.2 --- The Spatial-Temporal Relation Graph --- p.61 / Chapter 5.2.3 --- The Optimal Skim Problem --- p.62 / Chapter 5.3 --- Graph Optimization --- p.64 / Chapter 5.4 --- Static Video Summary Generation --- p.65 / Chapter 5.5 --- Experiments --- p.68 / Chapter 5.6 --- Summary --- p.74 / Chapter 6 --- Video Content Annotation and Semantic Video Summarization --- p.75 / Chapter 6.1 --- Semantic Video Content Annotation --- p.77 / Chapter 6.1.1 --- Video Shot Segmentation --- p.77 / Chapter 6.1.2 --- Semi-Automatic Video Shot Annotation --- p.77 / Chapter 6.2 --- Video Structures and Semantics --- p.78 / Chapter 6.2.1 --- Video Structure Analysis --- p.78 / Chapter 6.2.2 --- Video Structure and Video Edit Process --- p.80 / Chapter 6.2.3 --- Mutual Reinforcement and Semantic Video Shot Group Detection --- p.81 / Chapter 6.3 --- Semantic Video Summarization --- p.84 / Chapter 6.3.1 --- Summarization Requests and Goals --- p.84 / Chapter 6.3.2 --- Determine the Sub-Skimming Length for Each Scene --- p.85 / Chapter 6.3.3 --- Extracting Video Shots by String Analysis --- p.86 / Chapter 6.4 --- Experiments --- p.88 / Chapter 6.5 --- Summary --- p.92 / Chapter 7 --- Concluding Remarks --- p.93 / Chapter 7.1 --- Summary --- p.93 / Chapter 7.2 --- Future Work --- p.95 / Chapter A --- Notations --- p.97 / Bibliography --- p.100

Video segmentation is the first step to most content-based video analysis. In this thesis, several methods have been proposed to detect shot transitions including cut and wipe. In particular, a new cut detection method is proposed to apply multi-adaptive thresholds during three-step processing of frame-by-frame discontinuity values. A "likelihood value", which measures the possibility of the presence of a cut at each step of processing, is used to reduce the influence of threshold selection to the detection performance. A wipe detection algorithm is also proposed in our thesis to detect various wipe effects with accurate frame ranges. In the algorithm, we carefully model a wipe based on its properties and then use the model to remove possible confusion caused by motion or other transition effects. / With the segmented video shots, video indexing and retrieval systems retrieve video shots using shot-based similarity matching based on the features of shot key-frames. Most shot-based similarity matching methods focus on low-level features such as color and texture. Those methods are often not effective enough in video retrieval due to the large gap between semantic interpretation of videos and the low level features. In this thesis, we propose an attention-driven video retrieval method by using an efficient spatiotemporal attention detection framework. Within the framework, we propose an efficient method for focus of attention (FOA) detection which involves combining adaptively the spatial and motion attention to form an overall attention map. Without computing motion explicitly, it detects motion attention using the rank deficiency of gray scale gradient tensors. We also propose an attention-driven shot matching method using primarily FOA. The matching method boosts the attended regions in the respective shots by converting attention values to importance factors in the process of shot similarity matching. Experiment results demonstrate the advantages of the proposed method in shot similarity matching. / Li, Shan. / "September 2007." / Adviser: Moon-Chuen Lee. / Source: Dissertation Abstracts International, Volume: 69-02, Section: B, page: 1108. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2007. / Includes bibliographical references (p. 150-168). / 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. / Abstract in English and Chinese. / School code: 1307.

Ngai Li. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2001. / Includes bibliographical references (p. 75-[78]). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Mode of access: World Wide Web. / Abstracts in English and Chinese.

Zhang, Qian. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2011. / Includes bibliographical references (leaves 97-111). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstract also in Chinese.

Includes bibliographical references (leaves 199-214) / xxx, 214 leaves : ill. (some col.), plates (col.) ; 30 cm. / Title page, contents and abstract only. The complete thesis in print form is available from the University Library. / Thesis (Ph.D.)--University of Adelaide, Dept. of Electrical and Electronic Engineering, 2002

Until now the topic of motion estimation, as used in video compression, has been dominated by search methodologies which are modifications of an exhaustive search. This research takes a completely new approach by applying two evolvable search algorithms, the Genetic Algorithm and the Genetic Program, to the area of motion estimation. The main purpose of this research is to determine the applicability of evolvable search methods to the topic of motion estimation. Several methods are studied: in the first application, a Genetic Algorithm is used to determine individual motion vectors one at a time, while the second method explores the use of a Genetic Algorithm to search for all of the motion vectors to correlate two frames simultaneously. To reduce the number of motion vectors required, Genetic Programming is applied to variable block size motion estimation. Finally, this work is expanded by applying it to region motion estimation, which is not restricted to using square or rectangular motion blocks. / Graduation date: 1998

Thesis (Ph.D.)--Hong Kong University of Science and Technology, 2006. / Includes bibliographical references (leaves 130-138). Also available in electronic version.

This Bachelor thesis examines the possibility of replacing an outdated, analog video recording system to a digital counterpart. It is key that the video and audio signals remain synchronized, generator locked and time stamped. It is up to nine different video sources and a number of audio sources to be recorded and treated in such a manner which enables synchronized playback. The  different video sources do not always follow a universal standard, and differ from format as well as resolution. This thesis aims to compare a number of state of the art commercial of the shelf solutions with proprietary hardware. Great emphasis is placed on giving a functional view over the system features and to evaluate different compression methods. The report also discusses different transmission, storage and playback options. The report culminates in a series of proposed solutions to sub problems which are solved and treated separately, leading to a final proposal from the author. The final draft set how well the system meets pre-set requirements to price.