The primary objective of this thesis is to provide a low-complexity rate-distortion optimal coding mode selection method in digital video encoding. To achieve optimal compression efficiency in the rate-distortion framework with low computational complexity, we first propose a rate-distortion model and then apply it to the coding mode selection problem. The computational complexity of the proposed method is very low compared to overall encoder complexity because the proposed method uses simple image properties such as variance that can be obtained easily. Also, the proposed method gives significant PSNR gains over the mode selection scheme used in TM5 for MPEG-2 because the rate-distortion model considers rate constraints of each mode as well as distortion. We extend the model-based mode selection approach to motion vector selection for further improvement of the coding efficiency. In addition to our theoretical work, we present practical solutions to real-time implementation of encoder modules including our proposed mode selection method on digital signal processors. First, we investigate the features provided by most of the recent digital signal processors, for example, hierarchical memory structure and efficient data transfer between on-chip and off-chip memory, and then present practical approaches for real-time implementation of a video encoder system with efficient use of the features.

No description available.

No description available.

CoordenaÃÃo de AperfeÃoamento de Pessoal de NÃvel Superior / Ericsson Brasil / O aumento da demanda por serviÃos ricos em multimÃdia e a escassez do espectro eletromagnÃtico tÃm motivado a pesquisa de tecnologias capazes de aumentar a capacidade de sistemas sem fio sem requerer espectro adicional. Nesse contexto, comunicaÃÃes Dispositivo-a-Dispositivo (D2D, do inglÃs Device-to-Device) representam uma tecnologia promissora. Ao permitir comunicaÃÃo direta e de baixa potÃncia entre os dispositivos, comunicaÃÃes D2D levam a um maior e mais inteligente reuso dos recursos de rÃdio, permitindo um descongestionamento da rede de transporte de dados. Como resultado, a capacidade total do sistema e especialmente a eficiÃncia espectral sÃo aumentadas; e a proximidade entre os dispositivos permitem transferÃncias de dados com baixo atraso e altas taxas de dados, sem requerer potÃncia extra da bateria dos dispositivos. Entretanto, com o objetivo de tornar real os potenciais ganhos de comunicaÃÃes D2D como uma rede secundÃria da celular (primÃria), algumas questÃes chave precisam ser controladas. Assumindo que os dispositivos se comunicando estÃo cientes um do outro, a condiÃÃo do enlace (canal) deve ser avaliada. Caso seja benÃfica, tÃcnicas de GestÃo de recursos de rÃdio (RRM, do inglÃs Radio Resource Management) sÃo empregadas para que a interferÃncia co-canal causada nos dispositivos celulares seja mitigada. Tais tÃcnicas podem ser resumidas como: agrupamento, seleÃÃo de modo e controle de potÃncia. Nessa dissertaÃÃo, eu foco a minha atenÃÃo para RRM em comunicaÃÃes D2D subjacentes a redes LTE, e para as principais tÃcnicas de RRM para mitigar a interferÃncia co-canal. Objetivando a reduÃÃo da interferÃncia intra-celular e na melhoria da eficiÃncia espectral, eu formulo um problema conjunto de agrupamento e controle de potÃncia. Entretanto, devido à sua complexidade eu proponho mÃtodos sub-Ãtimos para agrupar usuÃrios celulares e D2D com o objetivo de minimizar a interferÃncia intra-celular, levando em conta a ortogonalidade espacial entre os usuÃrios que compartilham o recurso. AlÃm disso, eu analiso mÃtodos para decidir se um candidato D2D deveria se comunicar diretamente ou de modo convencional atravÃs da estaÃÃo rÃdio-base (eNB, do inglÃs Evolved Node B). Os resultados mostram que comunicaÃÃes D2D conseguem melhorar a eficiÃncia espectral do sistema e que a maioria dos ganhos pode ser alcanÃada agrupando de forma adequada os usuÃrios para compartilhar recursos baseando-se em projeÃÃes sucessivas e ortogonais, assim como combinando diferentes mÃtricas de compatibilidade espacial. AlÃm disso, nessa dissertaÃÃo eu argumento que tecnologias D2D podem ser usadas para aumentar ainda mais a eficiÃncia espectral e energÃtica se os parÃmetros chave dos algoritmos de RRM forem adequadamente estendidos para comunicaÃÃes D2D em mÃltiplos saltos. Especificamente, eu proponho um novo algoritmo distribuÃdo de controle de potÃncia baseado em maximizaÃÃo da utilidade que à capaz de equilibrar eficiÃncia espectral e energÃtica, enquanto leva em consideraÃÃo a seleÃÃo de modo e restriÃÃes na alocaÃÃo de recursos inerentes à integraÃÃo do ambiente celular-D2D. Os resultados numÃricos mostram que comunicaÃÃes D2D em mÃltiplos saltos combinadas com o algoritmo de controle de potÃncia proposto sÃo Ãteis nÃo apenas para colher os potenciais ganhos identificados na literatura, mas tambÃm para estender a cobertura de redes celulares. / The increasing demand for fast multimedia services and the scarcity of electromagnetic spectrum has motivated the research of technologies able to increase the capacity of wireless systems without requiring additional spectrum. In this context, Device-to-Device (D2D) communication represents a promising technology. By enabling direct and low-power communication among devices, D2D communication leads to an increased and intelligent spatial reuse of radio resources allowing to offload the data transport network. As a result, the overall system capacity and specially the spectral efficiency is increased; and the proximity between devices allows data transfer with low delays and high rates without requiring extra power from devicesâ batteries. However, in order to realize the potential gains of D2D communications as a secondary network of the cellular (primary) one, some key issues must be tackled. Assuming that the communicating devices are aware of each other, the actual link (channel) conditions must be evaluated. If beneficial, Radio Resource Management (RRM) techniques would be employed so that the co-channel interference caused in cellular devices would be mitigated. Such techniques may be summarized as: grouping, mode selection, and power control. In this thesis, I focus my attention on the RRM for D2D communications underlaying a Long Term Evolution (LTE)-like network, and the main RRM techniques to mitigate the co-channel interference. Aiming at the reduction of the intra-cell interference and at the improvement of spectral efficiency, I formulate a joint grouping and power allocation problem. However, due to its complexity I propose suboptimal methods to group cellular and D2D User Equipments (UEs) with the goal of minimizing intra-cell interference, taking into account spatial orthogonality between the UEs that share the same resources. In addition, I analyze methods to decide if D2D-capable UEs should communicate directly to one another or in the conventional way via the Evolved Node B (eNB). The results show that D2D communications can improve the spectral efficiency of the system and that most of this improvement can be achieved by suitably grouping the UEs for sharing resources based on successive orthogonal projections and matching different spatial compatibility metrics. Moreover, in this thesis I argue that D2D technology can be used to further increase the spectral and energy efficiency if the key D2D RRM algorithms are suitably extended to support network assisted multi-hop D2D communications. Specifically I propose a novel, distributed utility maximizing power control (PC) scheme that is able to balance spectral and energy efficiency while taking into account mode selection and resource allocation constraints that are important in the integrated cellular-D2D environment. The analysis and numerical results indicate that multi-hop D2D communications combined with the proposed PC scheme can be useful not only for harvesting the potential gains previously identified in the literature, but also for extending the coverage of cellular networks.

In order to extend the battery life of smart cameras, it is essential to increase the efficiency of power converters, especially at light load. This thesis research investigated a power converter to supply power for the microprocessor of a smart camera. The input voltage of the converter is 5 V, and the output voltage is 1.2 V with the load ranging from 10 mA (12 mW) to 1200 mA (1440 mW). The conventional buck converter is typically optimized for high efficiency at maximum load at the cost of light-load efficiency. A converter is investigated in this thesis to improve light load efficiency, while being able to handle heavy load, to prolong the battery life of smart cameras. The proposed converter employs two modes, a baby-buck mode and a heavy-load mode, in which each mode is optimized for its respective load range to achieve high efficiency throughout entire range. The heavy-load mode converter adopts the conventional synchronous buck approach, as it generally achieves high efficiency at heavy load. However, the synchronous buck approach is inefficient at light load due to the large switching, driving, and controller losses. The proposed baby-buck mode converter employs the following schemes or technique to reduce those losses. First, the baby buck mode converter adopts pulse frequency modulation (PFM) with discontinuous conduction mode (DCM) to lower the switching frequency at light load, so frequency-dependent switching and driving losses are reduced. Second, a simple control scheme, constant on-time V2 control, is adopted to simplify the controller and hence minimize the controller power dissipation. Third, the top switch of the baby-buck mode uses a small MOSFET, which is optimized for light load, and the bottom switch uses Schottky diode in lieu of a MOSFET to simplify the COT V2 controller. Fourth, the proposed converter combines the heavy-load and baby-buck mode converter into a single converter with a shared inductor, capacitors, and the feedback controller to save space. Finally, a simple and low power feedback controller with an optimum mode selector, a COT V 2 controller, and gate drivers are designed. The optimum mode selector selects an appropriate mode based on the load condition, while shutting down the opposing mode. The proposed converter was fabricated in CMOS 0.25 µm technology in two phases. Phase 1 contains design of the proposed converter with open loop, and its functionality is verified through measurements of test chips. Phase 2 includes the entire converter design with the feedback controller. Since the test chips of phase 2 are not delivered, yet, its functionality during the steady state and transient responses are verified through simulations. Simulation results indicate that the efficiency of the proposed converter ranges from 74% to 93% at 12 mW and 1440 mW, respectively. This result demonstrates that the proposed converter can achieve higher efficiency for the entire load range when compared to an off-the-shelf synchronous buck converters. / Master of Science

The Multiset Sampler [Leman et al., 2009] has previously been deployed and developed for efficient sampling from complex stochastic processes. We extend the sampler and the surrounding theory to model selection problems. In such problems efficient exploration of the model space becomes a challenge since independent and ad-hoc proposals might not be able to jointly propose multiple parameter sets which correctly explain a new pro- posed model. In order to overcome this we propose a multiset on the model space to en- able efficient exploration of multiple model modes with almost no tuning. The Multiset Model Selection (MSMS) framework is based on independent priors for the parameters and model indicators on variables. We show that posterior model probabilities can be easily obtained from multiset averaged posterior model probabilities in MSMS. We also obtain typical Bayesian model averaged estimates for the parameters from MSMS. We apply our algorithm to linear regression where it allows easy moves between parame- ter modes of different models, and in probit regression where it allows jumps between widely varying model specific covariance structures in the latent space of a hierarchical model. The Storytelling algorithm [Kumar et al., 2006] constructs stories by discovering and con- necting latent connections between documents in a network. Such automated algorithms often do not agree with user's mental map of the data. Hence systems that incorporate feedback through visual interaction from the user are of immediate importance. We pro- pose a visual analytic framework in which such interactions are naturally incorporated in to the existing Storytelling algorithm through a redefinition of the latent topic space used in the similarity measure of the network. The document network can be explored us- ing the newly learned normalized topic weights for each document. Hence our algorithm augments the limitations of human sensemaking capabilities in large document networks by providing a collaborative framework between the underlying model and the user. Our formulation of the problem is a supervised topic modeling problem where the supervi- sion is based on relationships imposed by the user as a set of inequalities derived from tolerances on edge costs from inverse shortest path problem. We show a probabilistic modeling of the relationships based on auxiliary variables and propose a Gibbs sampling based strategy. We provide detailed results from a simulated data and the Atlantic Storm data set. / Ph. D.

Video compression is a key technology used in several multimedia applications. Improvements in the compression techniques together with the increasing speed and optimized architecture of the new family processors enable us to use this technology more in real time systems. H.264 (also known as MPEG-4 Part 10 or AVC - Advanced Video Coding), is the latest video coding standard which is noted for achieving very high data compression. While H.264 is superior to its predecessors, it has a very high computational complexity which makes its costly for real time applications. Thus, in order to perform video encoding with satisfactory speed there is an obvious need for reducing the computational complexity. New algorithms were developed for this purpose. The developed algorithms were implemented on Texas Instrument TMS320C64x family to be able to fulfill the requirement in optimized signal processing hardware with low power consumption which arises from the computational complexity and the need for portable devices in video processing technology. With the new algorithms developed, a computation reduction of 55% was achieved without loosing perceptual image quality. Furthermore, the algorithms were implemented on a DSP along with the networking functionality to obtain a video streaming system. The final system may be used in a wide range of fields from surveillance systems to mobile systems.

<p>ITU - T and the Moving Picture Expert Group (MPEG) have jointly, under the name of Joint Video Team (JVT), developed a new video coding standard. The standard is called H.264 and is also known as Advanced Video Coding (AVC) or MPEG-4 part 10. Comparisons shows that H.264 greatly outperforms MPEG-2, currently used in DVD and digital TV. H.264 halves the bit rate with equal image quality. The great rate - distortion performance means nevertheless a high computational complexity. Especially on the encoder side.</p><p>Handling of audio and video, e.g. compressing and filtering, is quite complex and requires high performance hardware and software. A video encoder consists of a number of modules that find the best coding parameters. For each macroblock several $modes$ are evaluated in order to achieve optimal coding. The reference implementation of H.264 uses a brute force search for this mode selection which is extremely computational constraining. In order to perform video encoding with satisfactory speed there is an obvious need for reducing the amount of modes that are evaluated.</p><p>This thesis proposes an algorithm which reduces the number of modes and reference frames that are evaluated. The algorithm can be regulated in order to fulfill the demand on quality versus speed. Six times faster encoding can be obtained without loosing perceptual image quality. By allowing some quality degradation the encoding becomes up to 20 times faster.</p>

ITU - T and the Moving Picture Expert Group (MPEG) have jointly, under the name of Joint Video Team (JVT), developed a new video coding standard. The standard is called H.264 and is also known as Advanced Video Coding (AVC) or MPEG-4 part 10. Comparisons shows that H.264 greatly outperforms MPEG-2, currently used in DVD and digital TV. H.264 halves the bit rate with equal image quality. The great rate - distortion performance means nevertheless a high computational complexity. Especially on the encoder side. Handling of audio and video, e.g. compressing and filtering, is quite complex and requires high performance hardware and software. A video encoder consists of a number of modules that find the best coding parameters. For each macroblock several $modes$ are evaluated in order to achieve optimal coding. The reference implementation of H.264 uses a brute force search for this mode selection which is extremely computational constraining. In order to perform video encoding with satisfactory speed there is an obvious need for reducing the amount of modes that are evaluated. This thesis proposes an algorithm which reduces the number of modes and reference frames that are evaluated. The algorithm can be regulated in order to fulfill the demand on quality versus speed. Six times faster encoding can be obtained without loosing perceptual image quality. By allowing some quality degradation the encoding becomes up to 20 times faster.

The main purpose of the present study was to explore the effects of characteristics of information given in a job advertisement (amount of information and the specificity of the information) on the potential applicants&rsquo / willingness to apply to the job opening with the mediating roles of credibility of and satisfaction from the information, and attraction to the organization. In addition to that, the factors affecting the decision of the applicants to apply for the job posted such as application modes (internet, by hand in the same city the applicant lives, by hand in a different city than the applicant lives), different selection methods (interview, personality test, knowledge test), and personality characteristics (goal orientation and self-efficacy) were investigated. The study was conducted in three phases. In the first phase, one hundred and seven Middle East Technical University (METU) students were used to decide between two alternatives of goal orientation and two self-efficacy scales, which were translated into Turkish and adapted for the current study, to be used in the main study by determining the validity and reliabilities of the scales. The second phase was the manipulation check, conducted to test the clarity and meaningfulness of the newly developed job advertisements and questions related to the research hypotheses. Thirty-two METU students were used for the second phase. The final phase was the main study. One hundred and fifty four METU students from Electrical and Electronics Department were used to examine the hypotheses of the study.The students were randomly assigned to the three different versions of job advertisement. Additionally, a questionnaire package was given to each participant. The results supported most of the main hypotheses and the proposed model except for the mediating effect of satisfaction. Advertisement type affected the willingness to apply to the job of potential applicant and this relation was mediated by credibility and credibility was mediated by attractiveness to the organization. A main effect of application mode was found for change in application decision, but no interaction was found between application mode and different attraction levels. Goal orientation found to be affecting applicants&rsquo / decision change for different selection methods. Self-efficacy failed to predict decision change for application for different selection methods. Results are discussed along with the strengths and limitations of the study and suggestions for future research.