Group News
| 2009.06.05 | Direct Prediction for Subpel MV with Optimal MV Resolution Estimation by qi zhang (SAL 101 6/10/2009 1.00-2.00pm) |
|
Abstract: Abstract: A technique to predict the optimal sub-pel motion vector (MV) position based on integer-pel MVs is investigated in this work. Although it is possible to find the optimal MV position by fitting a local error surface using integer-pel MVs, the minimum of the error surface may fall outside of the region of interest. By analyzing the behavior of the error surface, we show that existing estimation scheme cannot accommodate increased subpel MV resolution without large estimation error. Hence, we propose a blockbased optimal subpel MV resolution estimation scheme that allows blocks with different characteristics to maximize its RD gain through a flexible MV resolution. Based on how well the error surface is conditioned, two different optimal MV prediction schemes are proposed: a novel direct optimal subpel MV prediction scheme is proposed for ill-conditioned blocks to estimate the optimal subpel position in one step without refining the resolution by half at a time. This direct prediction scheme is further extended to provide accurate prediction for well-conditioned blocks. Experimental results are given to show the low complexity and good RD performance of the proposed sub-pel MV prediction schem | |
| 2009.05.26 | Group Seminar Schedule for 2009 Summer and Fall, Version: 2 |
|
5/9-5/15 Early Summer Break 5/16-5/22 Early Summer Break 5/22-5/29 Dong Woo Kang 5/30-6/5 Tze-Ping Low 6/6-6/12 Sanjay Purushotham 6/13-6/19 Joyce Liang 6/20-6/26 Joohyun Cho 6/27-7/3 July 4th Holiday 7/4-7/10 Qi Zhang 7/11-7/17 Selina Chu --> Jessy Lee 7/18-7/25 Sachin Chachada 7/26-7/31 Roy Lou 8/1-8/7 Ping-Hao Wu 8/8-8/14 Late Summer Break 8/15-8/21 Late Summer Break 8/22-8/28 Late Summer Break 8/29-9/4 Jewon Kang 9/5-9/11 Yongjin Cho 9/12-9/18 Pei-Ying Chiang 9/19-9/25 Tanaphol Thaipanich 9/26-10/2 Seung-Hwan Kim 10/3-10/9 Chun-Su Park 10/10-10/16 Sungje Cho 10/17-10/23 Steve Cho 10/24-10/30 Youngmin Kwak 10/31-11/6 Sphinx Tsau 11/7-11/13 Jessy Lee --> Selina Chu 11/14-11/20 May-Chen Kuo 11/21-11/27 Thanksgiving 11/28-12/4 Shahryar Karimi 12/5-12/11 Winter Break 12/12-12/18 Winter Break 12/19-12/25 Winter Break 12/26-1/1 Winter Break 1/2-1/8 Winter Break 1/9-1/15 Jing Zhang 1/16-1/22 Jonghye Woo 1/23-1/29 Jasmine Wan | |
| 2009.06.23 | Secrecy analysis using two eavesdropper?s game by Joohyun Cho (SAL 101 6/26/2009 1.00-2.00pm) |
|
Abstract: Secrecy analysis is the recently emerging research field in wireless communication and network. Many researches related to secrecy analysis are based on Information Theory used for analyzing channel capacity. In addition, conventional research results have paid attention to secrecy capacity from the authentic users? point of view but, in this presentation, secrecy capacity is analyzed from the eavesdropper?s point of view and by two eavesdropper game. In two eavesdroppers? game, there are two eavesdroppers trying to catch authentic users? communication information and there is no central unit to manage eavesdropping. Under this condition, the target of this research is to find what the best strategy eavesdropper can take to achieve maximum information leakage capacity is and how their selfish behaviors converge to optimum strategy. Conventional wireless channel analysis, cooperative communication and game theory are applied to get the research result. The significance of this research is that we can measure the limit of Information leakage conducted by hostile users and make counter plan to prevent it | |
| 2009.06.17 | Cooperative Communications with Network Coding by Joyce Liang (SAL 101 6/19/2009 1.00-2.00pm) |
|
Abstract: The purpose of a communication network is to convey information from one device to another through what is often a noisy environment. The traditional network model assumes that nodes operate using store-and-forward routing over point-to-point links, yet this ignores what theory suggests: cooperation between nodes can increase information flow between source and destination. Within wireless networks, typical devices treat overhearing channels as noise and take countermeasures to mitigate this perceived interference. However, there are been numerous studies in strategies to process and relay the overheard transmission to benefit other nodes. Most efforts have been directed towards one-hop strategies. Network coding, is a rapidly growing research area that allows intermediate nodes to code outgoing messages based on incoming messages. Random linear network coding (RLNC) in particular restricts this operation to be a linear combination of randomly chosen coefficients but works well in a distributed manner with high probability. In this talk, we will discuss the fundamental principles in cooperative communication theory that lead up to network coding for simple networks using Shannon information theory. We will then discuss how RLNC can be a powerful tool for cooperative communication in a large distributed network | |
| 2009.06.08 | A Survey on Video Indexing and Retrieval by Sanjay Purushotham (SAL 101 6/12/2009 1.00-2.00pm) |
|
Abstract: Online Video repositories like YouTube, Google Videos etc., has added a new dimension in our entertainment life and has changed the way we access and view videos. The amount of video content available to a user on today's TV and Internet is enormous and the number of videos is exponentially growing every day. These videos are making our video repositories extremely huge and highly infeasible for humans to handle it manually. To find a video of interest for a user from these large video databases is a difficult task. Hence, there is a need to automatically classify and index videos for easy access and retrieval of the videos from these large databases. The goal of this talk is to introduce the challenges, approaches and open research problems in the field of video indexing and retrieval . In this talk, the findings of the past research, current trends and the future direction of this research field will be discussed. Content based video indexing, semantic-based video indexing and Multimodal video indexing will be presented | |
| 2009.06.01 | Optimized Opportunistic Multicast Scheduling (OMS) over Wireless Cellular Networks by Tze-Ping Low (SAL 101 6/5/2009 1.00-2.00pm) |
|
Abstract: Optimized opportunistic multicast scheduling (OMS) is studied for cellular networks, where the problem of efficiently transmitting a common set of fountain-encoded data from a single base station to multiple users over Rayleigh fading channels is examined. The proposed OMS scheme better balances the tradeoff between multiuser diversity and multicast gain by transmitting to a subset of users in each time slot using the maximal data rate that ensures successful decoding by these users. We first analyze the system delay in homogeneous networks by capitalizing on extreme value theory and derive the optimal selection ratio (i.e., the portion of users that are selected in each time slot) that minimizes the delay. Then, we extend results to heterogeneous networks where users are subject to different channel statistics. By partitioning users into multiple approximately homogeneous rings, we turn a heterogeneous network into a composite of smaller homogeneous networks and derive the optimal selection ratio for the heterogeneous network. Computer simulations confirm theoretical results and illustrate that the proposed OMS can achieve significant performance gains in both homogeneous and heterogeneous networks as compared with the conventional unicast and broadcast schedulin | |
| 2009.05.25 | Automatic heart segmentation in CT images by Dong-Woo Kang (SAL 101 5/29/2009 1.00-2.00pm) |
|
Abstract: Cardiovascular disease is the leading direct or contributing cause of non-accidental deaths in the world. Consequently, particular effort is being dedicated to early diagnosis and therapy. 3-dimensional imaging is now well established for noninvasive diagnosis of cardiovascular disease. Automatic image processing methods are a prerequisite to efficiently analyze the large amount of image data produced by computed tomography (CT) scanners during cardiac exams. Compared with other imaging modalities, cardiac CT can provide detailed anatomic information about the heart chambers, large vessels, and coronary arteries. There are major tasks to develop such an automatic segmentation system: heart modeling (shape representation) and automatic model fitting (detection and segmentation). Due to the complexity of cardiac anatomy, it is not trivial to represent the anatomy accurately while keeping the model simple enough for automatic segmentation and manual correction if necessary. In this talk, I will introduce the segmentation in medical image processing, and discuss the future research on automatic heart segmentation | |
| 2009.05.12 | Group Seminar Schedule for 2009 Summer and Fall |
|
5/9-5/15 Early Summer Break 5/16-5/22 Early Summer Break 5/22-5/29 1 Dong Woo Kang 5/30-6/5 2 Tze-Ping Low 6/6-6/12 3 Sanjay Purushotham 6/13-6/19 4 Joyce Liang 6/20-6/26 5 Joohyun Cho 6/27-7/3 July 4th Holiday 7/4-7/10 6 Qi Zhang 7/11-7/17 7 Selina Chu 7/18-7/25 8 Sachin Chachada 7/26-7/31 9 Roy Lou 8/1-8/7 10 Ping-Hao Wu 8/8-8/14 Late Summer Break 8/15-8/21 Late Summer Break 8/22-8/28 Late Summer Break 8/29-9/4 1 Jewon Kang 9/5-9/11 2 Yongjin Cho 9/12-9/18 3 Pei-Ying Chiang 9/19-9/25 4 Tanaphol Thaipanich 9/26-10/2 5 Seung-Hwan Kim 10/3-10/9 6 Chun-Su Park 10/10-10/16 7 Sungje Cho 10/17-10/23 8 Steve Cho 10/24-10/30 9 Youngmin Kwak 10/31-11/6 10 Sphinx Tsau 11/7-11/13 11 Jessy Lee 11/14-11/20 12 May-Chen Kuo 11/21-11/27 Thanksgiving 11/28-12/4 13 Shahryar Karimi 12/5-12/11 Winter Break 12/12-12/18 Winter Break 12/19-12/25 Winter Break 12/26-1/1 Winter Break 1/2-1/8 Winter Break 1/9-1/15 1 Jing Zhang 1/16-1/22 2 Jonghye Woo 1/23-1/29 3 Jasmine Wan | |
| 2009.05.04 | Alternative Splicing Regulation Mechanism - from evolutionary point of view by Jing Zhang (THH 110 5/8/2009 1.00-2.00pm) |
|
Abstract: Alternative splicing is the RNA splicing variation mechanism in which the exons of the pre-mRNA, are separated and reconnected to produce alternative ribonucleotide arrangements. These linear combinations then undergo the process of translation into a greater variety of proteins. Alternative splicing is of great importance to genetics ? On one hand, it has been proposed that for eukaryotes alternative splicing was a very important step towards higher efficiency, because information can be stored much more economically. On the other hand, it allowed the new proteins much faster than in prokaryotes. Therefore, understanding the regulation mechanism of alternative splicing is becoming a hot topic these days. Cooperatively acting pairs of cis-regulatory elements play important roles in many biological processes. Here, we use a statistical approach to analyze pairs of markers that preferentially co-evolve in some specific area, which might take charge of controlling splicing events. Ongoing research and preliminary results will be discussed and vision for future work will be in | |
| 2009.04.20 | Texture Processing for Image/Video Coding and Super-Resolution Applications by Byung Tae Oh (THH 110 4/24/2009 1.00-2.00pm) |
|
Abstract: Textured image/video processing for coding and super-resolution is presented in this talk to improve coding efficiency and prediction accuracy. The research consists of three main parts. First, a synthesis-based texture coding technique that uses low-quality video as the side information to control the output texture for video coding is proposed. As compared with the current pure synthesis algorithm, the proposed algorithm is generic, in the sense that the behavior and quality of the output texture can be adjusted by the amount of the side information and determined by the user. We develop an area-adaptive side information assignment technique to improve coding efficiency by given bit-budget. Additionally, we also provide the texture decomposition algorithm to maximize the synthesis performance by decomposing the non-synthesizable illumination component from the input video. Second, a new technique for film grain noise extraction, modeling and synthesis is proposed and applied to high definition video coding. In this research, we extract film grain noise from the input video as a pre-processing step (at the encoder) and re-synthesize the film grain noise and add it back to the decoded video as a post-processing step (at the decoder). Under this framework, the coding gain of the denoised video is higher while the quality of the final reconstructed video is still high. To implement such a scheme, we develop a technique to remove film grain noise from image/video without distorting its original content as well as a parametric model consisting of a small set of parameters to represent the extracted film grain noise. This model generates film grain noise close to the real one in terms of power spectral density and cross-channel spectral correlation. Third, a texture interpolation technique based on the locally piecewise auto-regressive (PAR) model and the non-local (NL) training procedure is investigated. The proposed PAR/NL scheme selects model parameters adaptively based on local image properties with an objective to improve the interpolation performance of non-adaptive models, e.g., the bicubic algorithm. To determine model parameters for stochastic texture, we use the non-local (NL) learning algorithm to update and refine these local model parameters under the assumption that the PAR model parameters are self-regular. As compared to previous interpolation algorithms, the proposed PAR/NL scheme boosts texture details, and eliminates blurring artifacts perceptually | |
| 2009.04.13 | Diffusion MRI by Shahryar Karimi-Ashtiani (THH 110 4/17/2009 1.00-2.00pm) |
|
Abstract: Diffusion MRI opened a new front for uncovering the convoluted structure of the central nervous system by providing the capability for non-invasive identifications of geometries of white tracts in the brain. It is well understood that, this imaging modality is characterized by the shape of the self-diffusion profile within the brain fibers. Despite previous efforts in the literature for quantification of this physical phenomenon, the current methods suffer from a number of constraints which severely limit the extent of their practical applicability. In this work, we address the solution of Partial Differential Equation of self-diffusion in its most general case, by relaxing the previous work constraints. To this end, we developed an approach based on the Finite Elements methodology to discover the solution of self-diffusion in multi-compartments models of white tracts. For the finite element discretization of the problem, tetrahedral higher-order elements alongside the Method of Lines are utilized. Due to finite resolution of D-MRI signals, reconstruction of the data is voxel-wise, rather than point-wise. As a result, the D-MRI signal is sensitive to the average self-diffusion profiles of individual voxels regions. We formulated this problem in terms of parameters of microstructures of the white tracts, passing through a voxel. Consequently, the developed method can easily accommodate challenging situations such as tract crossings and demyelinations, in computation of the voxels average propagators | |
| 2009.04.06 | Compression Schemes for Motion Capture Data by May-Chen Kuo (THH 110 4/10/2009 1.00-2.00pm) |
|
Abstract: Motion Graph (MG) based motion synthesis are widely used for generating human motion in many applications, such as movie production, game simulation, military training, etc. The quality of the motion generated by the MG based approaches depends on the richness of the motion capture (mocap) database. Fortunately, the mocap technology advances a lot in the past decade. The cost of the mocap hardware drops down to be affordable for the universities and research labs. It is easy to collect a huge amount of mocap data. Mocap data can be stored in terms of captured marker positions, or degree of freedom (dof) with respect to the captured skeleton. Considering the data captured a generic human model with 60 dofs at 120 fps presented in single precision (four bytes), each second of motion is of about 30KB, i.e., a one hour-long motion is of 100MB. This constrains many applications from using a large database even though it is available and promises better quality. For example, in the scenario of on-line games, the network bandwidth is limited such that it is against the transmission of large volume of data. A compression algorithm is thus desired as the hope to untangle this dilemma. The compression mechanisms for videos are extensively explored in the past decades. However, those well developed scheme are not suitable for the mocap data for two main reasons. First, the frames are no longer 2D images, thus the JPEG scheme does not apply. Second, the naturalness of human motion also lies in some high frequency information, such as footprints, so they should be properly preserved. The compression schemes for 3D animated meshes are not optimal for the mocap data, because that in a motion, each frame can be of different temporal importance and each dof (vertex in 3D animated meshes) can of different spatial significance. We design our algorithm to take advantage of the correlation in time and in space for human motion. The compression gain is three-fold. One is temporal down-sampling to preserve the keyframes only. Another is spatial down-sampling to preserve the key dofs only. The last is the quantization on the timing and positional inform | |
| 2009.04.02 | Recognition and characterization of unstructured environmental sounds by Selina Chu (THH 110 4/3/2009 1.00-2.00pm) |
|
Abstract: Recognizing environmental sounds is a basic audio signal processing problem. Consider, for example, applications in robotic navigation, assistive robotics, and other mobile device-based services, where context aware processing is often desired. Human beings utilize both vision and hearing to navigate and respond to their surroundings, a capability still quite limited in machine processing. The first step toward achieving multi-modality is the ability to process unstructured audio and recognize audio scenes (or environments). The goal of this thesis is on the characterization of unstructured environmental sounds for understanding and predicting the context surrounding of an agent or device. Most research on audio recognition has focused primarily on speech and music. Less attention has been paid to the challenges and opportunities for using audio to characterize unstructured audio. My research will investigate challenging issues in characterizing unstructured environmental sounds such as the development of appropriate feature extraction algorithm and learning techniques for modeling the variations of the environment | |
| 2009.03.25 | HHT-based MP decomposition by En-Shuo Tsau (THH 110 3/27/2009 1.00-2.00pm) |
|
Abstract: Matching Pursuit is a well-known greedy algorithm to find a sub-optimal solution to the problem of an adaptive approximation of a signal in a redundant set (dictionary) of functions, i.e. Gabor atoms, which was first proposed in 1993 by Dr, Mallat. The MP provides an simple and easy implement way to the problem; however, the major obstacle of MP is the running complexity is very high provided the huge overcomplete basis. HHT was first introduced by Dr. Huang in 1998 using the principle of empirical mode decomposition (EMD) and the Hilbert transform. The EMD process conducts a nonlinear approach to first decompose the signal into several intrinsic mode functions (IMFs) and their instantenous frequency. In our algorithm, we try to adopt HHT as a preprocessing of MP and provide a nice estimator of Gabor's atom parameters to archive a fast version MP. In the seminar, I will introduce our algorithm and some of the experiments will be presented | |
| 2009.03.09 | Hybrid Lattice Boltzmann Method by Youngmin Kwak (THH 110 3/13/2009 1.00-2.00pm) |
|
Abstract: The fluid simulation is the core of special effects in computer animations, computer games, and movies. The lattice Boltzmann method numerically solves the Boltzmann equations and enables a fast simulation and visualizing splashing effect of fluid. In this seminar, hybrid Lattice Boltzmann method will be proposed and some interesting simulation results will be visualized | |
| 2009.03.03 | Steganalysis Performance Estimation for Large Image Database by Steve Cho (THH 110 2/6/2009 1.00-2.00pm) |
|
Abstract: Steganalysis is the art and science of detecting messages hidden using steganography. In blind steganalysis, a blind detector learns what a typical cover image looks like in a multi-dimensional feature space and a classifier is then trained to learn the differences between cover and stego image features. I will introduce some blind steganalysis methods using Wavelet-based, DCT-based, and Markov-based features. Instead of using classifiers in feature space, we are trying to apply hypothesis testing framework to evaluate detection performance. In addition, I would like to focus on selection of image database (image characteristics, number of images), which can be influential in detection performance of steganalysis | |
| 2009.02.24 | Coronary CT angiography (CTA) with multi-slice helical scanners by Jonghye Woo (THH 110 2/27/2009 1.00-2.00pm) |
|
Abstract: Coronary CT angiography (CTA) with multi-slice helical scanners is becoming the integral part of major diagnostic pathways for coronary artery disease. In addition, coronary CTA has demonstrated substantial potential in quantitative coronary plaque characterization. If serial comparisons of plaque progression or regression are to be made, accurate 3D volume registration of these volumes would be particularly useful. In this talk, we present feature-based coronary CTA volume registration of the paired coronary CTA scans. We achieve this with a combined registration strategy, which uses the global rigid registration as an initialization, followed by local registration using non-rigid volume registration with a volume-preserving constraint. We exploit the extracted coronary trees to help localize and emphasize the region of interest. The computational efficiency and robustness of the proposed registration have been demonstrated by the experiments of real patient dat | |
| 2009.02.11 | Overview of 3D shape Retrieval by Pei-Ying Chiang (THH 110 2/13/2009 1.00-2.00pm) |
|
Abstract: Recent developments in techniques for modeling, digitizing and visualizing 3D shapes has led to an explosion in the number of available 3D models on the Internet and in domain-specific databases. This has led to the development of 3D shape retrieval systems that, given a query object, retrieve similar 3D objects. In this presentation, I am going to give an overview of content based 3D retrieval. Start with an introduction of the shape analysis, the shape analysis challenging, proposed solutions will be presented later. The solutions are evaluated with respect to several requirements of content based 3D shape retrieval, such as: (1) shape representation requirements, (2) properties of dissimilarity measures, (3) efficiency, (4) discrimination abilities, (5) ability to perform partial matching, (6) robustness, and (7) necessity of pose normalization. Finally, the conclusion and future directions are discussed | |
| 2009.02.02 | Underdetermined Audio Source Separation in Reverberant Environments by Namgook Cho (THH 110 2/6/2009 1.00-2.00pm) |
|
Abstract: We propose a technique to separate audio sources from their anechoic mixtures with long delay in an underdetermined setting, i.e., the number of audio sensors is smaller than that of sources). It consists of two stages: 1) to estimate anechoic mixing parameters of attenuation and arrival delay and 2) to recover original audio sources based on estimated mixing parameters. When delay is longer than one sample, previous algorithms perform poorly. To address this shortcoming, we estimate the maximum delay and use it to find a proper frequency range that produces no phase ambiguity. Then, we determine mixing parameters with time-frequency points in this range. Finally, mathematical tools are used to solve the underdetermined linear system to recover original audio sources. The effectiveness of the proposed technique on various mixing scenarios with noisy observation of mixtures and different types of sounds is demonstrated by experimental results | |
| 2009.01.27 | Generalization of Benford's Law based on Statistical Distributions for Image Forensics by Jingwei Wang (THH 110 1/30/2009 1.00-2.00pm) |
|
Abstract: We analyze a generalization of Benford's law and investigate the possible extensions of Benford's law for image forensic applications in this paper. This generalization is based on several different statistical distribution models. We show that the proposed Generalization is more suitable to image statistical models than the Benford's law. We perform exemplary simulations to point out image forensic applications of the proposed model, e.g., double JPEG compression and dithering. The results prove our proposed model is effective and reasonable | |
| 2009.01.20 | Efficient Management Techniques for Large Video Collection by Ping-Hao Wu (THH 110 1/23/2009 1.00-2.00pm) |
|
Abstract: In this research, we focus on two techniques related to the management of large video collection: video copy detection and automatic video classification. We propose a fast duplicate video detection system using the suffix array data structure in this work. To achieve this objective, we first develop a compact one-dimensional signature based on the underlying video structure, which is discriminative yet insensitive to various attacks. Being motivated by the genome sequence alignment, we adopt an efficient matching technique using the suffix array data structure. The proposed system can perform the sequence matching in linear time while the complexity of conventional duplicate video copy detection algorithms grows at least quadratically with the video length. Two novel features that take the shooting process into consideration are proposed for video genre classification. Although a lot of work has been proposed with the consideration of cinematic principles, most features are still extracted at lower levels without much semantic information. We propose two features that estimate the number of camera used and the distribution of the camera distance in a short time interval, respectively. Preliminary experiment results show that the proposed features capture additional genre-related information. Some conclusion about the genre can be inferred from the proposed features to a certain degree. This is still an on-going research topic | |
| 2009.01.12 | Dependent R-D modeling for H.264/SVC bit allocation by Yongjin Cho (THH 110 1/16/2009 1.00-2.00pm) |
|
Abstract: In this presentation, we propose bit allocation algorithms for H.264/SVC based on dependent analytical R-D models. With H.264/SVC videos, the requirement for the scalability support makes the inter-dependency relation among coding units becomes even more complex than that with H.264/AVC videos. For this reason, it becomes very important to develop bit allocation algorithms based on the knowledge of dependent R-D characteristics of coding units. The dependency consideration in the video coding often involves prohibitive complexity, which prevents them from their practical employment. To address the complexity issue, we propose dependent R-D models for two scalable dimensions of the temporal and the quality scalability in this work. Then, temporal layer and joint quality-temporal layer bit allocation algorithms are developed based on the proposed dependent R-D models. The performance of the proposed bit allocation algorithms is verified by the comparison with the JSVM benchmark, where significant coding efficiency enhancement could be observed | |
| 2008.01.01 | 2009 Spring Seminar Schedule |
|
12/6-12/12 Winter break 12/13-12/19 Winter break 12/20-12/26 Winter break 12/27-1/2 Winter break 1/3-1/9 Winter break 1/10-1/16 1 Ping-Hao Wu 1/17-1/23 2 Yongjin Cho 1/25-1/30 3 Pei-Ying Chang 1/31-2/6 4 Namgook Cho 2/7-2/13 5 Jingwei Wang 2/14-2/20 6 Sungje Cho 2/21-2/27 7 Jonghye Woo 2/28-3/6 8 Steve Cho 3/7-3/13 9 Youngmin Kwak 3/14-3/20 Spring Break 3/21-3/27 10 Sphinx Tsau 3/28-4/3 11 Jessy Lee 4/4-4/10 12 May-Chen Kuo 4/11-4/17 13 Shahryar Karimi-Ashtiani 4/18-4/24 14 Byung Tae Oh 4/25-5/1 15 Paul Cha 5/2-5/8 16 Jing Zhang 5/9-5/15 Early Summer Break 5/16-5/22 Early Summer Break 5/22-5/29 1 Dong Woo Kang 5/30-6/5 2 Tze-Ping Low 6/6-6/12 3 Sanjay Purushotham 6/13-6/19 4 Joyce Liang 6/20-6/26 5 Joohyun Cho 6/27-7/3 6 Qi Zhang 7/4-7/10 7 Angel Dai 7/11-7/17 8 Selina Ch | |
| 2008.12.04 | Secrecy Analysis and Information by Joohyun Cho (VHE 217 12/5/2008 1.00-2.00pm) |
|
Abstract: Secrecy analysis is the recently emerging research field in wireless communication. The main interest of secrecy analysis is how much information can be transmitted securely in various communication environments such as ad-hoc network, broadcasting network, co-operative communication and etc. The first step into secrecy analysis is to understand ?what is perfect secrecy?? In this presentation, I will introduce the concept of classical perfect secrecy defined by Claude E. Shannon and extend this to the recent wireless communication issues with several examples and explain how secrecy capacity is related with information leakage and game th | |
| 2008.11.20 | Transfer learning for acoustic environment classification by Selina Chu (VHE 217 11/21/2008 1.00-2.00pm) |
|
Abstract: Real-world environments are dynamic and unpredictable, making the audio data likely to be different, but sharing similar properties. When working with unstructured data, such as those of environment sound, we are constantly faced with many different and unlabeled data, which makes learning a model with good generalization difficult using standard supervised learning approaches. Since unlabeled data is relatively easy to collect, but lacks ways to use them, we could leverage them by using semi-supervised learning to build better classifiers. One type of semi-superivsed algorithm is transfer learning, which uses the knowledge acquired in a set of source tasks to improve performance in a related target task. In this talk, I will begin by presenting a brief introduction on semi-supervised and transfer learning and describe a possible way of doing implementing this concept. I will conclude by discussing the applicability of this work to environmental sounds | |
| 2008.11.12 | Overview of Multi-core Power-aware Video Coding by Roy Lou (VHE 217 11/14/2008 1.00-2.00pm) |
|
Abstract: With continuously increasing demand of computation, the thirst of high performance processor has never stopped. However, the technology of semiconductor is approaching the limit that it cannot catch up with Moore's Law. In order to meet the computation requirement, multi-core processors has become popular recently. On the other hand, as the emerging of mobile computation, handheld devices, like cell phone and laptop, are widely-spread as comsumer product, while battery technology still left far behind. In this talk, we will have an overview of two hot topic of computer engineering - multi-core and power-aware. Related video application will be also discussed. Finally, conclusion and future direction will be presented | |
| 2008.11.02 | Texture Super-resolution by Byung Tae Oh (VHE 217 11/7/2008 1.00-2.00pm) |
|
Abstract: Increasing the resolution of the image have been paid attention during last two decades, and it is still interesting and challenging problem for both research and practical system. In this talk, introduction of the general background knowledge about the super-resolution will be presented first, and we will explore various super-resolution algorithms to see its advantages and disadvantages. Next, we will narrow our focus down to texture as one of the specific image contents, and introduce the proposed texture interpolation algorithm. Experimental results will be also given to show the performance of the proposed algorithm with comparing to other methods. Finally, the conclusion and the future direction of the proposed algorithm will be presented | |
| 2008.10.28 | Cascaded Intra Prediction for High Definition Image/Video Coding by Yunyang Dai (VHE 217 10/31/2008 1.00-2.00pm) |
|
Abstract: Due to the rising demand for high fidelity and super resolution contents, the current coding trends are gearing towards high fidelity image/video coding. In this work, a novel block-based intra-prediction scheme is proposed for efficient image (or intra-frame) coding, where we apply various 2D geometrical manipulations to reference image blocks to enrich the pool of prediction blocks for a given target block. As compared with the traditional line-based intra prediction in H.264/AVC, the new scheme offers a significant coding gain (about 0.24-1.23dB in the PSNR value with the same bit rate) at the cost of higher complexity. An enhanced 2DIP scheme (2DEIP) is further proposed to reduce the complexity and enhance the efficiency of 2DIP for images with complex textures. Finally a cascaded intra prediction (CIP) is proposed to apply intra prediction in a systematical way which incorporates both the simplicity of LIP and the boosted efficiency from 2DEIP with the cascaded approach. Experimental results and conclusions will be given in the end | |
| 2008.10.21 | Enhanced residual coding for lossless and near-lossless high definition video content by Qi Zhang (VHE 217 10/24/2008 1.00-2.00pm) |
|
Abstract: In this talk, we will first introduce the limitations of current coding standards regarding lossless and near lossless coding. We will examine H.264 AVC standard?s performance in lossless and near lossless modes. We will compare the performance of H.264 with other state of art lossless coding standards such as JPEG-LS in order to gain insights regarding their respective strengths and weaknesses. We will show current standards are inefficient in dealing with lossless and near-lossless content coding. The presence of granular noise in the residual of the high definition content renders the current standards ineffective. We will propose an enhanced scheme that can be used as a direct add-on of the current coding standards or as a part of a new scalable codec. The proposed algorithm uses pixel domain compensation in dealing with lossless coding and frequency domain direct compensation for near-lossless coding. Experimental result will be reported for both the lossless and near-lossless case | |
| 2008.10.13 | Cryptography, A Survey into its Past, Present, and Future by Joyce Lian (VHE 217 10/17/2008 1.00-2.00pm) |
|
Abstract: The term cryptographia was first coined in the 17th century and is widely known through popular culture and daily use in modern applications such as secure e-mail and e-commerce. However, cryptography began almost as early as the written language, and its current manifestation exhibits remarkable similarities to its earliest forms. The goal of this talk is to provide a window into its history and how it has helped to shape modern civilization. Cryptanalysis has naturally developed as the antagonist in this story to break what is thought to be unbreakable. Largely thanks to Claude Shannon, however, we now have a metric for security and a condition for perfect security. We will examine his metric and use it to prove the impenetrable secrecy of a some secret sharing schemes and in particular, Network Coding | |
| 2008.10.05 | MC-CDMA-based fingerprinting and its new results by Byung-Ho Paul Cha (VHE 217 10/10/2008 1.00-2.00pm) |
|
Abstract: Multiple copies of media contents can be delivered to end users through multi-casting network. A traitor tracing is an important application to identify the unauthorized usage of distributed contents on multi-casting network. A collusion attack is powerful in breaking the traitor tracing system, and a time-varying collusion attack is easy to apply and difficult to detect. A fingerprinting system robust to time-varying collusion attacks has never been studied in the past. We have developed multi-carrier code division multiple access (MC-CDMA)-based fingerprinting system to resolve this problem. In this talk, we characterize multimedia, time-varying collusion attacks, and user as well as colluder capacity to clarify the property of multimedia fingerprinting systems. We first introduce important components for multimedia fingerprinting system. Second, we provide more detailed analysis for time-varying collusion attacks depending on variations of colluder weight on colluder and sample index. Third, we derive multimedia characteristics by multimedia representation for both music and movie with human perceptual system models. Then, we connect multimedia characteristics with bandwidth and power-limited situation, which is analogous to the finite resource environment in a wireless communication. We also provide user and pilot code allocation schemes as the extension of colluder weight estimation (CWE) technique, and we introduce pilot codes with time and frequency domain considerations. Finally, we derive the relationship between the number of identified colluders and fingerprint-to-interference and noise ratio (FINR). Through those characterizations, we provide more clear understanding of proposed MC-CDMA-based fingerprinting system and more possibility to develop much richer and advanced fingerprinting system | |
| 2008.09.29 | Diffusion MRI Technology in Exploring the Brain Microstructures by Shahryar Karimi-Ashtiani (VHE 217 10/3/2008 1.00-2.00pm) |
|
Abstract: In this talk, I'll go over the different major aspects of diffusion MRI technology and the applications to unveiling the brain tissue structures. I will open with a brief overview of the Central Nervous System (CNS) in human, and move onto diffusion process and its significance in exploring biological tissues. I will have a short discussion regarding the principles of diffusion MRI, and different quantification approaches, including DTI, for assessment of the brain structures. The importance of the validation problem will be highlighted and the necessity for having a gold standard to evaluate different tractography algorithms will be discussed. At the end, our proposed method for development of a diffusion MRI simulator will be explained | |
| 2008.09.21 | Motion Capture Data Compression by May-Chen Kuo (VHE 217 9/26/2008 1.00-2.00pm) |
|
Abstract: The quality of the motion synthesis depends on the richness of the motion capture (mocap) database. However, the file size is an issue - a one hour-long single character motion is of 100MB. This constrains many applications from using a large database even though it is available and promises better quality. For example, in the scenario of on-line games, the network bandwidth is limited such that it is against the transmission of large volume of data. A compression algorithm is thus desired to untangle this dilemma. The compression mechanisms for videos are extensively explored in the past decades. However, those well developed scheme for the mocap data for two main reasons. First, the frames are no longer 2D images, thus the JPEG scheme does not apply. Second, the naturalness of human motion also lies in some high frequency information, such as footprints, so they should be properly preserved. The compression schemes for 3D animated meshes are not optimal for the mocap data, because that in a motion, each frame can be of different temporal importance and each dof (vertex in 3D animated meshes) can of different spatial significance. We design an algorithm to take advantage of the correlation in time and in space tailored for motion captured data. The mocap data is composed of the value of each degree of freedom (dof) over time. The compression gain is from approximating the curve of each dof by consecutive unimodal functions. If the captured object is in contact with the environment, such as a character standing on the ground, this approximation will lead to visual artifacts at the contact point, such as the foot sliding or penetrating the ground. This artifact is known as foot-skating artifact in the domain of computer animation. To fix this, the timing and the positions of the contact points are also compressed as the guide to remove the foot-skating artifact after decompression. The compression and decompression can be processed faster than realtime (30 fps). Our algorithm has higher compression ratio with little quality degradation as compared to the existing approaches | |
| 2008.09.16 | Simplification and perceptual analysis of 3D tree models by Jessy Lee (VHE 217 9/19/2008 1.00-2.00pm) |
|
Abstract: The effectiveness of 3D tree model simplification techniques and performance metrics are presented in this talk. Previous work had investigated simplification techniques based on pixel-based metrics that did not take the tree model's leaf density in consideration. We group leaf objects in a tree model and perform simplification based on the density of how the final rendered image would appear. Objective and subject performance analyses are conducted to evaluate how well the algorithms are able to produce tree models that appear as close to the original tree model. Rather than comparing how many pixels are different among simplified and full tree model, we propose to use Gabor filter for analyzing the orientation and spatial relationship within the rendered tree models. In addition, psychophysical evaluation is conducted by asking a group of observers to rate how close each simplified model resembles the original model. Experimental results show that objective metric and psychophysical experiment reach conclusions that are consistent with each other | |
| 2008.09.11 | Congratulations to Usman Riaz for passing her Defense exam! |
|
Congratulations to Usman Riaz for passing his Defense on September 11. The title of his thesis is "Coding and Optimization for Ultra-Wide-Band Wireless Communications". His PhD thesis committee includes: Jay Kuo (Chair), Solomon W Golomb and Remigijus Mikulevicius (Outside Member) | |
| 2008.09.10 | Congratulations to Youngmin Kwak for passing his Qual exam! |
|
Congratulations to Youngmin Kwak for passing his Qual today. The title of his thesis proposal is "Advanced Liquid Simulation Techniques for Computer Graphic Applications". His thesis guidance committee includes: Jay Kuo (Chair), Antonio Ortega, Jerry Mendel, Shri Narayanan and Ulrich Neumann (Outside Member) | |
| 2008.09.09 | Congratulations to Ronald Chang for passing her Defense exam! |
|
Congratulations to Yu-Jung (Ronald) Chang for passing his Defense today (with some minor revision required for his thesis completion). The title of his thesis is "Resource Allocation in OFDM/OFDMA Cellular Networks: Protocol Design and Performance Analysis". His thesis committee includes: Jay Kuo (Chair), Mike Neely and Ramesh Govindan (Outside Member). Wish him the very best in his job hunting | |
| 2008.09.09 | HHT-based multipitch analysis in polyphonic music by Sphinx Tsau (VHE 217 9/12/2008 1.00-2.00pm) |
|
Abstract: Multiple fundamental frequency tracking has always been tough in musical information retrieval task. Unlike conventional approach, we try to utilized HHT as a tool to develop a robust detection scheme in polyphonic case. HHT was first introduced by Dr. Huang in 1998 using the principle of empirical mode decomposition (EMD) and the Hilbert transform. The EMD process conducts a nonlinear approach to first decompose the signal into several intrinsic mode functions (IMFs), which provide multi-scale in time domain. HHT is successful in dealing with non-stationary and non-linear data and has been applied to several area like mechanical engineering, aerospace engineering. It has several benefit like no time-frequency resolution trade-off comparing to Fourier Transform, better visualization. However, it still has several constrains for newly development. In the presentation, first I will introduce HHT including its insight, pros and cons. Second, i will discuss multipitch analysis in polyphonic music, the convention method, the challenge and the links with HHT. Several experiments will be provided. Finally i will make a conclusion and future work | |
| 2008.09.08 | Congratulations to Jason Chen for passing her Defense exam! |
|
Congratulations to Jason Chen for passing his defense on September 8 (Monday). The title of his thesis is "Robust Transmit Precoding and V-BLAST Receiver Design for MIMO Communication Systems". His thesis defense committee includes: Jay Kuo (Chair), Xiaoli Yu and Ming-Deh Huang (Outside Member) | |
| 2008.09.01 | Advanced Liquid Simulation Techniques for Computer Graphics Applications by Youngmin Kwak (VHE 217 9/5/2008 1.00-2.00pm) |
|
Abstract: The fluid simulation is the core of special effects in computer animations, computer games, and movies. The particle level set method numerically solves the Navier Stokes equations and is a hybrid method which combines the level set and particles to enable a realistic and smooth fluid (liquid) simulation. The lattice Boltzmann method numerically solves the Boltzmann equations and enables a fast simulation and visualizing splashing effect of fluid. In this seminar, the advanced particle level set method and hybrid Lattice Boltzmann method will be proposed and some interesting simulation results will be visualized | |
| 2008.08.04 | Hybrid Steganography by Steve Cho (SAL 150 8/8/2008 1.00-2.00pm) |
|
Abstract: Steganography is the art and science of writing hidden messages in such a way that no one apart from the sender and intended recipient even realizes there is a hidden message. First, I will start the introduction part with going over basic concept and keyword in this field. Application and future of stegangraphy will also be discussed. Steganographic algorithms write hidden messages to make communication undetectable. Several steganographic algorithms like F5, OutGuess, and Model-based will be introduced. Steganalytic methods detect the presence of hidden data. In blind steganalysis, a blind detector learns what a typical cover image looks like in a multi-dimensional feature space. Several steganalytic methods using wavelet, DCT, and Markov-based features will be examined. Finally, I would like to propose ?hybrid steganography?, which takes advantage of different steganographic methods. I will explain basic concept of this method and show some preliminary experimental results | |
| 2008.07.28 | Curve Evolution with A Dual Shape Similarity and Its Application to Segmentation of Left Ventricle by Jonghye Woo (SAL 150 8/1/2008 1.00-2.00pm) |
|
Abstract: Segmentation has been playing a critical role in medical image analysis and Level set methods have shown an efficacy in various imaging modalities. In this talk, we present a novel segmentation approach to jointly delineate the boundaries of epi- and endocardium of the left ventricle on the Magnetic Resonance Imaging (MRI) images in a variational framework using level sets, which is in great demand as clinical applications in cardiology. One strategy to tackle segmentation under undesirable conditions such as subtle boundaries and occlusions is to exploit prior knowledge which is specific to the object to segment, in this case the knowledge about heart anatomy. While most left ventricle segmentation approaches incorporate a shape prior obtained by a training process from an ensemble of examples, we exploit a novel shape constraint using an implicit shape prior knowledge, which assumes shape similarity between epi- and endocardium allowing a variation under the Gaussian distribution. Our approach does not demand a training procedure which is usually subject to the training examples and is also laborious and time-consuming in generating the shape prior. Instead, we model a shape constraint by a statistical distance between the shape of epi- and endocardium employing signed distance functions. We applied this technique to cardiac MRI data with quantitative evaluations performed on 10 subjects. The experimental results show the robustness and effectiveness of our shape constraint within a Mumford-Shah segmentation model in the segmentation of left ventricle from cardiac MRI images in comparison with the manual segmentation results | |
| 2008.07.20 | Underdetermined Anechoic Audio Source Separation by Namgook Cho (SAL 150 7/25/2008 1.00-2.00pm) |
|
Abstract We consider the problem of underdetermined blind audio source separation of anechoic mixtures. First, we investigate the limitations and basic assumption of the previous work. To overcome the limitations and to deal with more realistic situations, we propose a two-stage demixing technique that exploits the sparsity of source signals. In the first stage, the mixing parameters are estimated in the anechoic scenario, where both attenuations and arrival delays between sensors are considered. Then, in the second stage, we use the estimated mixing parameters to extract the source signals from the mixtures by solving the underdetermined system that has fewer observation microphones than signal sources. To the end, the generalized lp-norm minimization is employed for the blind source extraction stage. Experimental results on various scenarios with speech, music, and environmental sounds demonstrate significant gains in separation performance when compared to the previous work | |
| 2008.07.07 | DNA sequence assembly with short fragments by Sungje Cho (SAL 150 7/11/2008 1.00-2.00pm) |
|
Abstract To sequence a human genome takes too much money, but our current goal is $1,000 genome for a human. Nowadays, to achieve this goal, we are focusing on high throughput sequencer which produces reads very fast, but produces very short reads. It means that the high throughput sequencer can reduce the cost of sequencing, but it makes assembler too difficult to assemble those fragments. Recently the Eulerian path method is proposed for it. As the reads length getting shorter and needs for analyzing indel mutations getting more demanding, we believe that there can be only one solution for this framework; it is the Eulerian graph method. I will shortly state the history of DNA sequencing, and propose our idea in this talk | |
| 2008.06.22 | Is there any thing left in advanced video coding? by C.-C. Jay Kuo (SAL 150 6/27/2008 1.00-2.00pm) |
|
Abstract There has been a significant progress in video coding techniques in the last two decades. Many standards have been established along the way, including MPEG-1, MPEG-2, MPEG-4, H.261, H.263 and H.264. It is fair to say that the video coding research field has reached a certain degree of maturity, and ?Is there anything left for video coding?? has been often asked. There is no easy answer to this question. There has been some preliminary effort to explore the possibility of a new video coding standard, H.265, with an objective to reduce the bit rate by 50% while keeping video quality about the same. The slow start of the H.265 standard activity reflects the difficulty of identifying major break-through video coding technologies. However, from the market viewpoint, there is still a strong driving force to reach a higher coding gain especially for high definition (HD) video content. There exist quite a few differences between HD and SD (simple definition) contents, which could be potentially exploited. Several new research directions will be pointed out in this talk. It includes: multi-view video coding, joint encoder-decoder optimization based on the context information, analysis-by-synthesis (or model-based) texture coding, new distortion measurement tools, etc. These new research directions will be discussed in this talk | |
| 2008.06.18 | Motion recognition with Wii Remote by PeiYing Chiang (SAL 150 6/20/2008 1.00-2.00pm) |
|
Abstract Wii has become the most successful game product within these years;the intuitive user interaction design makes the user learn to play games easily. The motion sensor, three axes accelerometer, which is built inside of the Wii remote (Wiimote) can detect the accelerating and tilt along three dimension. With the IR sensor, the Wiimote can detect the relative translation as well. A great diversity of new game design has revealed because of the game controller revolution. There are more and more developers try to propose the innovative applications on their game or mobile device platforms with this low cost motion sensor.However, the motion sensor also has limitations which make it difficult to extend its application variety. In this prepensentation, we are going to discuss about current motion recognition technologies, motion analysis Wiimote, and how we overcome the weakness to benefit future application | |
| 2008.06.08 | A Camera Motion-driven Visual Saliency Map Model for Videos by Shilin Xu (SAL 150 6/13/2008 1.00-2.00pm) |
|
Abstract Visual saliency map regards to the extraction of Region-of-Interest (ROI) from videos or images. How to automatically model the ROI for a given video is the main problem to be solved in this topic. Since each video varies from others in many basic features such as motion, texture, color, intensity, object shape, etc, it is very challenging to design a reliable computational ROI model. However, for videos, the motion plays a critical role in visual saliency map because motion can easily attract our attention and then trigger the sequential observation and tracking of HVS (Human visual system). Therefore, extracting the moving objects in the foreground is main concern of our research. Considering the camera is always moving with the shooting targets, the camera motion should be evaluated before foreground extraction. We propose a modified version of camera motion estimation algorithm and apply it to extract the foreground objects. Also, we use other methods, such as background modeling and contrast map, to form our final saliency map model. In the presentation, I will show some interesting results and propose my future work | |
| 2008.06.01 | Bit Allocation for Spatial Scalability in H.264/SVC by Jiaying (Shirley) Liu (SAL 150 6/6/2008 1.00-2.00pm) |
|
Abstract H.264/SVC is recently standardized as a scalable extension of H.264/AVC. The production of R-D optimized bit stream is one of major objectives in the video encoder design. In this presentation, we will discuss about a model-based spatial layer bit allocation algorithm for H.264/SVC. The spatial scalability of H.264/SVC is achieved by a multi-layer approach, where an enhancement layer is bound by the dependency on its preceding layers. The inter-layer dependency is decoupled in our analysis by a careful examination of the signal flow in the encoder. We show that the rate and the distortion characteristics of a dependent layer can be represented by a number of independent functions with a group of pictures (GOP) as a basic coding unit. Finally, a low complexity spatial layer bit allocation scheme is developed using the proposed GOP-based R-D models | |
| 2008.05.30 | Seamless High-Velocity Handover Support in Mobile WiMAX Networks by Zhiwei YAN (SAL 150 5/30/2008 1.00-2.00pm) |
|
Abstract The IEEE 802.16e standard (mobile WiMAX) has been proposed to provide connectivity to wireless networks for mobile users (including users at the vehicular speed). It is first shown in our analysis that the handover successful probability decreases significantly when the user moves at a higher speed. Then, we propose the use of the forward error correction (FEC) to offer extra protection of handover signaling messages to enhance the handover successful probability, especially at a higher velocity. Moreover, an adaptive FEC scheme can be adopted to achieve a target handover successful probability at various velocities. It is demonstrated by simulation results that the proposed FEC scheme achieves the target handover successful probability efficiently | |
| 2008.05.15 | Congratulations to Gracia Wang for passing her Qual exam! |
|
Congratulations to Gracia Wang for passing her Qual exam today. The title of her thesis proposal is: "Robust Video Transmission in Erasure Networks with Network Coding". Her thesis committee includes: Jay Kuo (Chair), Zhen Zhang, Bhaskar Krishnamachari, Panayiotis Georgiou and Cyrus Shahabi (Outside Member) | |
| 2008.05.12 | Congratulations to Wesley Lee for passing her Defense exam! |
|
Congratulations to Wesley Lee for passing his defense today. His PhD thesis title is "H.264/AVC Decoder Complexity Modeling and Its Applications". His thesis guidance committee includes: Jay Kuo (Chair), Mihaela van der Schaar and Tu-Nan Chang | |
| 2008.05.06 | Fast Duplicate Video Detection Using Suffix Array Data Structure by Ping-Hao Wu (ZHS 163 5/9/2008 1.00-2.00pm) |
|
Abstract We propose a fast duplicate video detection system using the suffix array data structure in this work. To achieve this objective, we first develop a compact one-dimensional signature based on the underlying video structure, which is discriminative yet insensitive to various attacks. Unlike images and audio, the size of videos is usually very large, which makes it computationally expensive to match two very long signature sequences. Being motivated by the genome sequence alignment, we adopt an efficient matching technique using the suffix array data structure. The proposed system can perform the sequence matching in linear time while the complexity of conventional duplicate video copy detection algorithms grows at least quadratically with the video length | |
| 2008.04.30 | Robust Multimedia Transmission in Erasure Network with Network Coding by Gracia Wang (ZHS 163 5/2/2008 1.00-2.00pm) |
|
Abstract Network coding techniques have been proposed for efficient data multicast over networks. Practical network coding adopts random linear network coding method which combines a group of packets by randomly selecting the coefficient on a finite field in a distributed way. However, the loss of one random linear encoded packet means the lost of a dimension for the constrained equations in decoding. Unless global network coding matrix is of full rank, we are not able to recover all source packets by error concealment. Specifically, we have addressed two significant topics: 1) Construct a sparse global coding matrix for layered H.264/SVC (scalable video coding) bit-stream transmission and maintain the shape of the sparse matrix through the random linear network coding process. The graceful quality degradation is achieved by partial decoding of the received data block. 2) Propose an interleaving scheme to enable network coding and error concealment cooperating effectively. Moreover, the unequal erasure protection bed on the scheme achieves the smooth quality degradation | |
| 2008.04.20 | Synthesis-based texture coding for video compression by Byung Tae (ZHS 163 4/25/2008 1.00-2.00pm) |
|
Abstract The synthesis-based texture coding technique is investigated in this talk to improve the coding efficiency of the current video coding standard. First, e will breifly introduce the current texture synthesis techniques and synthesis-based texture coding method. Next, a synthesis-based texture coding technique that uses low-quality video as the side information to control the output texture for video coding is proposed. Furthermore, an area-adaptive side information assignment scheme is developed to improve coding efficiency. To evaluate the efficiency of the proposed technique, we study a quantitative quality measure for the synthesized texture, which even provides an explicit criterion whether the input texture is synthesizable. Computer simulation results are given to demonstrate the superior performance of the proposed technique | |
| 2008.04.14 | Rate control of H.264/SVC by Yongjin Cho (ZHS 163 4/16/2008 1.00-2.00pm) |
|
Abstract H.264/SVC is recently standardized as a scalable extension of H.264/AVC. The production of an R-D optimized bit stream is one of the major objectives in the video encoder design. In this presentation, we will discuss about a model based temporal layer bit allocation scheme as a solution to the R-D optimal video encoding problem. We have three parts in the presentation. In the first part, we provide a brief overview of H.264/SVC along with possible research issues and major challenges. A dependent distortion model is proposed in the second part. In the R-D modeling of video signals, the inter-layer dependency is one of the major obstacles. With the proposed distortion model, the dependency issue could be resolved successfully. Finally, in the third part, we introduce a temporal layer bit allocation scheme as an application of the distortion model | |
| 2008.04.09 | Congratulations to Byung Tae for passing his Qual exam! |
|
Congratulations to Byung Tae for passing his Qual exam today. His thesis proposal title is "Synthesis-based Texture Coding for Video Compression". His thesis guidance committee includes: Jay Kuo (Chair), Sanjit Mitra, Jerry Mendel, Krishna Nayak and Ulrich Neumann (Outside Member) | |
| 2008.04.09 | Multi-carrier Collusion-Resistant Code (Robust Fingerprinting) by Joohyun (Peter) Cho (ZHS 163 4/11/2008 1.00-2.00pm) |
|
Abstract The collusion-resistant fingerprinting system for traitor tracing applications using a multi-carrier approach is described in this talk. One key issue in the fingerprinting system design is its robustness against the collusion attack. We focus on time-invariant and time-variant collusion attacks depending on the variations of colluder weights. Collusion-resistance can be achieved by choosing Hadamard-Walsh (HW) spreading codes followed by the multi-carrier modulation. For this reason, our hiding codes are called multi-carrier collusion-resistant codes (MCRC). Based on our recent development of fingerprinting system, we design and analyze the MCRC system against the time-invariant and time-variant colluder weights. For time-invariant colluder weights, we propose a colluder detection scheme with advanced user detectors to identify a larger number of colluders in weighted collusion attacks. The task of advanced user detection and colluder weight estimation can be formulated as a single-user and multi-user detection problem aided by channel estimation module from communication system. Based on multi-carrier code division multiaccess (MC-CDMA) system, our advanced detection scheme includes the maximal ratio combining (MRC) scheme for single-user detection in the same group and the parallel interference cancellation (PIC) multiuser detection scheme for multi-user detection in different groups. For time-variant colluder weights, the allocation schemes of pilot symbols and user codes with repetition coding are proposed, which are named as space-time spreading (STS) MCRC. The STS MCRC system allocates pilot symbols and user codes with repetition into the entire time or space of multimedia, and the system can be robust against the variations of colluder weights | |
| 2008.04.03 | Trust Without Privacy: Security Considerations in Mobile Ad Hoc Networks (MANETs) by Joyce Liang (ZHS 163 4/4/2008 1.00-2.00pm) |
|
Abstract The next generation of wireless communications will emphasize the rapid deployment of networks that enable survivable, efficient, and dynamic communications, i.e. mobile ad-hoc networks (MANETs). MANETs find particular relevance in military applications, as many governments shift towards a policy of cooperative communications in an infrastructureless, coalition environment. However, because the limited transmission range requires multiple hops in the path, each node in a MANET must act as source, intermediate, and receiver node. Network Coding (NC) has been shown as an attractive alternative to classical routing to achieve increased throughput, yet because of the open medium, it is still vulnerable to internal eavesdroppers. The tradeoff between security and efficiency can be jointly considered in the concept of trust. We suggest a model for the calculation of trust between heterogeneous nodes, which can be used for achieving information security when combined with NC in a wireless MANET | |
| 2008.04.03 | Game Theory and its applications on wireless network by Joohyun (Peter) Cho (ZHS 163 4/4/2008 1.00-2.00pm) |
|
Abstract Game Theory is a mathematical tool developed to understand competitive situations in which rational decision makers interact to achieve their objectives. Since the distinguished paper "Theory of Games and Economic Behavior" by John von Neumann and Oskar Morgenstern, game theory has been mainly studied and developed in economics. However, its characteristic, providing a mathematical tool for analyzing strategic situation, makes its applications more expanded into various areas such as biology, computer science, politics, and philosophy. In addition, game theory techniques have recently have been applied to various engineering design problems in which the action of one component impact (and perhaps conflicts with) that of any other component. In particular, game theory techniques have been successfully used for protocol design and optimization (e.g. radio resource management, power control) in wireless networks. In this presentation, I introduce game theory concepts with some examples and shortly state a survey on the game theory based resource management and admission control schemes in different wireless networks and several open research issues are outlined. Then game theory application on cognitive radio is presented | |
| 2008.03.13 | Relay-Assisted Decorrelating Multiuser Detector (RAD-MUD) for Cooperative CDMA Networks by Wan-Jen (Athena) Huang (ZHS 163 3/14/2008 1.00-2.00pm) |
|
Abstract In this talk, we consider the uplink of a cooperative CDMA network, where users cooperate by relaying each other's messages to the base station. When spreading waveforms are not orthogonal, multiple access interference (MAI) exists at the relays and the destination, causing cooperative diversity gains to diminish. To address this problem, we propose the relay-assisted decorrelating multiuser detector (RAD-MUD) to separate interfering signals at the destination with the help of precoding at relays along with pre-whitening at the destination. Unlike the conventional zero-forcing (ZF)precoder or the decorrelating MUD, the proposed RAD-MUD experiences neither power expansion at the relays nor noise amplification at the destination. Three cooperative transmission strategies are considered on top of RAD-MUD;namely, transmit beamforming, selective relaying and distributed space-time coding. Since the reliability of each source-relay and/or relay-destination links are different,relay transmissions are weighted accordingly in our schemes to further combat MAI.The advantages of RAD-MUD over existing cooperative MUD schemes are verified by computer simulation | |
| 2008.03.05 | Motion Synthesis with Wavelet-based Heterogeneous Motion Graph by May-chen Kuo (ZHS 163 3/7/2008 1.00-2.00pm) |
|
Abstract Heterogeneous Motion Graph (HMG) contains heterogeneous nodes and links. Each node is characterized by the wavelet transformed representation of a group of similar motion of a limb. A directed transitable link connects one node to another if they are transitable. A compatibility link connects two nodes if their associated limbs can be joined together. A walk in the graph along the compatibility links through nodes of different tiers leads to a pose. New poses which do not exist in the original database can be synthesized by HMG, and so can the new transitions. The feasibility of the synthesized poses and the related motion are verified with rules defined in bio-mechanical literatur | |
| 2008.02.27 | HHT-based multipitch analysis in polyphonic music by EnShuo (Sphinx) Tsau (ZHS 163 2/29/2008 1.00-2.00pm) |
|
Abstract Multiple fundamental frequency tracking has always been tough in musical information retrieval task. Unlike conventional approach, we try to utilized HHT as a tool to develop a robust detection scheme in polyphonic case. HHT was first introduced by Dr. Huang in 1998 using the principle of empirical mode decomposition (EMD) and the Hilbert transform. The EMD process conducts a nonlinear approach to first decompose the signal into several intrinsic mode functions (IMFs), which provide multi-scale in time domain. HHT is successful in dealing with non-stationary and non-linear data and has been applied to several area like mechanical engineering, aerospace engineering. It has several benefit like no time-frequency resolution trade-off comparing to Fourier Transform, better visualization. However, it still has several constrains for newly development. In the presentation, first I will introduce HHT including its insight, pros and cons. Second, i will discuss multipitch analysis in polyphonic music, the convention method, the challenge and the links with HHT. Several experiments will be provided. Finally i will make a conclusion and future work | |
| 2008.02.22 | Quantization Index Modulation (QIM) for Digital Watermarking by SeongHo (Steve) Cho (ZHS 163 2/22/2008 1.00-2.00pm) |
|
Abstract Many information-embedding algorithms like low-bit(s) modulation (LBM) method and additive spread-spectrum-based method have been proposed in the past. However, both LBM-type strategies and additive spread spectrum (SS) are, in general, not good choices for most information embedding and digital watermarking applications. A new class of information-embedding strategies called ˇ§Quantization Index Modulationˇ¨ (QIM) has been proved to be robust against arbitrary bounded and fully informed attacks, which arise in several copyright applications. In particular, QIM is proven to achieve provably better rate distortionˇVrobustness tradeoffs than currently popular spread-spectrum and low-bit(s) modulation methods. Before talking about QIM, I will give general background of digital rights management (DRM) and watermarking. After these general concepts, I will compare several information embedding methods (LBM, SS, QIM) for watermarking and their examples. QIM will be explained with detailed implementation example. Several extension of QIM like distortion-compensated QIM (DC-QIM) and spread transform dither modulation (STDM) will also be discussed. Finally, application of QIM to video watermarking will be considered | |
| 2008.02.12 | Variational approach and its applications by Jonghye Woo (ZHS 163 2/15/2008 1.00-2.00pm) |
|
Abstract We can interprete same phenomonen with different point of view such as analytic or statistical framework. Each framework has its strengths and weaknesses and can be applid to the different applications. Variational approach is widely used technique in the image processing and computer vision area such as image denoising, segmentation, registration, various inverse problems and super-resolution. A non-rigid ultrasound image registration method is proposed in this work using the intensity as well as the local phase information under a variational framework. One application of this technique is to register two consecutive images in an ultrasound image sequence. Although intensity is the most widely used feature in traditional ultrasound image registration algorithms, speckle noise and lower image resolution make the registration process difficult. By integrating the intensity and the local phase information, we can find and track the non-rigid transformation of each pixel under diffeomorphism between the source and target images. Experiments using synthetic and in vivo mice cardiac images are conducted to demonstrate the advantages of the proposed method | |
| 2008.02.07 | Group seminar: DNA sequence assembly with short fragments by Sungje Cho (ZHS 163 2/8/2008 1.00-2.00pm) |
|
Abstract After Sanger and Gilbert invented a DNA sequencing method, there was great development to sequence DNA, assemble it, and analyze the human DNA sequence. Though there were many researches about DNA sequencing methods, it costs too much yet. First human DNA sequence was sequenced with $3 billion, and today it costs several million dollars. But our goal is $1,000 genome for each person. Nowadays, to achieve this goal, we are focusing on high throughput sequencer which produces sequences very fast, but produces very short fragments. It means that the high throughput sequencer will reduce the cost of sequencing, but it makes assembler too difficult to assemble those fragments. Recently proposed idea is Eulerian path (or equivalently cycle) method to conquer this. Eulerian path finding problem is historically very old problem. It was suggested to help Sequence by Hybridization (SBH) sequencing method first in sequence assembly area. After that, SBH technology was disappeared without practical application. But as the fragment size is getting shorter, Eulerian path approach is re-visiting to assemble with very short fragments. I will shortly state the history of DNA sequencing, and propose our idea in this talk | |
| 2008.02.07 | Congratulations to Byung-Ho (Paul) Cha for passing his Qual exam! |
|
Congratulations to Byung-Ho (Paul) Cha for passing his Qual exam (1/29/2008). His thesis proposal is entitled with "Design of Collusion-Resistant Fingerprinting Systems with a Multicarrier Approach". His thesis guidance committee includes: Jay Kuo (Chairperson), Sanjit K. Mitra, Shri Narayanan, Richard Leahy and Ming-Deh Huang (Outside Member) | |
| 2008.01.28 | Congratulations to Namgook Cho for passing his Qual exam! |
|
Congratulations to Namgook Cho for passing his Qual exam 1/28/2007. The title of his thesis proposal is "Theory and Applications of Efficient Musical Signal Representation". His thesis committee includes: Jay Kuo (Chairperson), Jerry Mendel, Richard Leahy, Shri Narayanan and Liang Chen (Outside Member) | |
| 2008.01.28 | Group seminar: Disparity Estimation from Multi-View Images and Video: Graph Models and Algorithms by Jong Dae Oh (ZHS 163 2/1/2008 1.00-2.00pm) |
|
Abstract In this work, we consider the problem of estimating the depth information from the following three scenarios: a stereo image pair, multi-view images, and stereo image sequences. For stereo image matching, we first propose a new disparity map estimation algorithm for slant and curved surfaces. In this situation, we focus on two properties of the disparity map. The first one is continuous disparity change inside an object while the second one is sharp disparity change between object boundaries. To exploit these two properties at the same time, two techniques are proposed to improve the performance of existing stereo matching algorithms. To address disparity discontinuity in object boundaries, we present a disparity estimation procedure, which consists of two steps: a greedy disparity filling algorithm and a least-squared-errors (LSE) fitting method. Furthermore, it is observed that the existing fronto-parallel model with color segmentation is built upon the piecewise constant surface approximation. This is however not efficient in approximating slanted or curved objects. We propose to use a piecewise linear surface model to represent 3-dimensional (3D) geometric structure for better surface modeling. The proposed stereo matching system with these two new components is evaluated with Middlebury data sets with excellent quantitative and qualitative results. Then, a new graph model for disparity estimation of multi-view images is investigated. Two performance metrics for algorithmic evaluation are considered. They are quality and complexity. To reduce complexity, a graph model whose computational time is independent of the number of input images is proposed. With this model, new smoothness and consistency terms in the energy function are added to maintain the quality of a disparity map. Experimental results show that the proposed algorithm offers good enough quality at a much lower complexity than existing methods. To further improve the quality of the disparity map in stereo matching, we propose an advanced framework based on new graph models. Along this direction, we divide stereo matching problem into 3 sub-problems: 1) disparity estimation for non-occlusion regions and occlusion detection, 2) disparity estimation for occlusion regions, and 3) post-processing of the disparity map. A three-step procedure is proposed to solve them sequentially. At the first step, we perform an initial matching and develop a new graph model using the ordering constraint to improve disparity values in non-occlusion regions and detect occlusion regions. At the second step, we determine disparity values in occlusion regions based on global optimization. Since the conventional segmentation-based stereo matching is not efficient in highly slanted or curved objects, we propose a post-processing technique for disparity map enhancement based on a 3-dimensional (3-D) geometric structure. The proposed three-step stereo matching procedure yields excellent quantitative and qualitative results with Middlebury data sets. Finally, disparity estimation and virtual view synthesis from stereo video inputs are examined. To enhance the overall performance, a two-stage algorithm for accurate and fast disparity estimation and occlusion handling is first presented. Then, a new virtual view synthesis method with a preprocessing algorithm is described. The preprocessing algorithm can remove false matched regions for disparity refinement effectively. This synthesis method can reduce the blurring and ghostly effects and provides an excellent tradeoff in terms of computational time and synthesized video quality | |
| 2008.01.25 | Congratulations to Wan-Jen (Athena) Huang for passing his defense! |
|
Congratulations to Wan-Jen (Athena) Huang for passing his defense on 1/23/2008. Athena's thesis title is "Design and Analysis of High-Performance Cooperative Relay Networks". Her thesis guidance committee includes: Jay Kuo (Chairperson), Michael Neely, Xiaoli Yu and Fengzhu Sun (Outside Member) | |
| 2008.01.25 | Group seminar: Error concealment in scalable video coding (SVC) by Tanaphol Thaipanich (ZHS 163 1/25/2008 1.00-2.00pm) |
|
Abstract: H.264/SVC (Scalable Video Coding) is recently emerging standard. It allows a much more flexible bit-steam which is highly adaptable to bandwidth and hardware restriction. It has a decent coding efficiency via a development of H.264. However, as of now, there is very little research study on error concealment algorithm in SVC and most of current error concealment algorithms are not designed to exploit new features offered in SVC. Additionally, some concepts of available EC algorithms might not be able to function well in SVC without modification. In this week presentation, I will start my talk with a general overview of error concealment concept and introduction to SVC. After that we will discuss with related works and some of my experiment results so that we can have a better understanding of the subject matter. Then, I will present my current work and our progress. The conclusion will be given at the end to summary my researches | |
| 2008.01.17 | Group seminar: Improved Coding Scheme with Enlarged Block Size on High Definition Video Contents by Qi Zhang (ZHS 163 1/18/2008 1.00-2.00pm) |
|
Abstract: High definition video content is one of the most important consumer electronics trends developed in the past several years. High definition video content has several unique characteristics comparing to previous standard video content. First is the content fidelity. Very fine details needs to be faithfully reproduced in order to satisfy costumer demand. Second, the resolution is typically much larger than standard size video. The current HD resolution is about 1920x1080, even higher resolution is being introduced into the market such as the upcoming digital cimena 2K and 4K standard. Studies on HD contents point out that due to the increased resolution, the average block texture complexity is often much simpler than that of the standard definition (SD) contents such as the CIF resolution.Therefore, previous researches proposed to use larger macroblock size to provide better coding performance. However, experiment results show that the coding gain from using larger macroblock size is unstable amongst sequences with different complexity. In addition, coding gain from larger block size diminishes as the quantization parameter decreases. In this paper, we will provide a systematic analysis on the unique characteristics of HD contents. We identified the two major issues that prohibit efficient coding using larger block size on HD contents are film grain noise and gradient change. To combat these problems, we propose a pyramid based encoding scheme combined with fractal based solutions to enable the use of larger macroblock size. Experimental results will be given to show the effectiveness of our algorithm | |
| 2007.12.04 | Group seminar: Content-based Video Copy Detection by Ping-hao Wu (WPH 207 12/7/2007 1.00-2.00pm) |
|
Abstract: With the proliferation of digital video content on the internet and broadcasting, how to detect duplicate videos becomes a critical issue. Unlike watermarking, which inserts a distinct pattern into the video, video copy detection relies on the content-based signatures of the videos. The main challenges are the robustness of the features extracted and the massive capacity of video sources. In this presentation, I will begin by introducing the general concept of video copy detection and reviewing several existing methods. By relating the video matching problem to the sequence alignment problem, an approach that can quickly scan through a large video database to find duplicate of a given query video will be presented. Experiment results of testing several query videos against an 100 hour database will be given. Then I will conclude by discussing the plan for future works | |
| 2007.12.04 | Congratulations to Bei Wang for passing his defense! |
|
Congratulations to Bei Wang for passing his defense on 12/4/2007. Her thesis title is "Techniques for Efficient Cloud Modeling, Simulation and Rendering". Her thesis committee consists of Jay Kuo (Chair), Krishna Nayak and Ulrich Neumann (Outside Member) | |
| 2007.11.15 | Congratulations to Chia-Chun Hsu (Alex) for passing his defense! |
|
Congratulations to Chia-Chun Hsu (Alex) for passing his defense on 11/13/2007. His PhD thesis topic is "Coexistence Mechanisms for Legacy and Next Generation Wireless Networks Protocols".His Thesis Committee includes C.-C. Jay Kuo (Chair), Michael Neely and Ramesh Govindan (Outside Member) | |
| 2007.11.13 | Group seminar: Novel Intra Prediction Techniques for HD Content Coding by Yunyang Dai (WPH 207 11/16/2007 1.00-2.00pm) |
|
Abstract: With the evolution of digital consumer electronics and the increased capacity of network infrastructure, consumers are no longer satisfied with contents of coarse quality and limited resolution. Today?s sophisticated consumer increasingly demands high quality contents. Fidelity has become one of the main concerns regarding digital content delivery. There have been different attempts from both academic and industry community to support the rising demand for higher resolution with better visual quality. For example, JPEG2000 and H.264 all intra coding were incorporated in DCI for better coding quality and individual frame accessibility. Although the current intra prediction techniques used in H.264 is very efficient and shows some ability to capture the surface structure of the predicted image block, the residual image block still shows regular structures after the spatial prediction. In this talk, we will first present a novel intra prediction for high definition video coding. Different from the traditional line prediction techniques, we use 2D surface based prediction to further enhance the prediction quality. Experiment results will be presented to show the effectiveness of our algorithm. We will then present additional enhancements in fractal dimensions that allow further performance gain | |
| 2007.11.01 | Seminar: Paradigm Shifts in Modern IT Era Driven by Digital Media and Broadband Networks by Professor C.-C. Jay Kuo (WPH 207 11/2/2007 2.00-3.00pm) |
|
Information technologies have become more mature after its quick development for almost half a century. We are in a new era in which IT research has shifted its focus from technology development to novel applications. For example, digital technologies have been widely applied to speech, audio, video and graphics in various commercial applications today. Furthermore, the availability of broadband wired/wireless Internet infrastructures and new technologies such as peer-to-peer networking has changed the way of digital media distribution and exchange. This lecture will address the following four major IT paradigm shifts: - From the analog implementation to the digital implementation - From PC-centric to network-centric - From one-way broadcasting to two-way interaction - From HW/SW/infra-structure provision to contents and value-added service | |
| 2007.10.26 | Seminar: Overview of SIGINT Operations and Research Opportunities by Bill Nead (EEB 248 10/26/2007 2.00-3.00pm) |
|
Abstract: The main purpose of this talk is to optimize university research and assist with establishing an academic outreach program to best support the war fighters' needs on preparing the Battle space, Multi-INT collection management, processing techniques, and reporting. Visit introduction will include a detailed overview of the Intelligence Community as it relates to Multi-INT operations and how SIGINT Operations play a critical role | |
| 2007.10.24 | Group seminar: Background Modeling for Acoustic Environments by Selina Chu (WPH 207 10/26/2007 1.00-2.00pm) |
|
Abstract: Detecting foreground events from background audio in an environment is an important problem in an automatic acoustic scene recognition system. Background modeling is aimed at discriminating the expected information, namely the background from the unexpected events, i.e., the foreground. In general, most of the work in this area has been at the visual level. However, using vision alone has its limitation. Audio data could be obtained at any moment when the system is functioning, in spite of challenging external conditions such as poor lighting or visual obstruction, and is relatively cheap to store and compute compared to visual signals. Since other modality information are easily available (e.g., audio), it could effectively be used to discover unusual ˇ§activity patternˇ¨ in a scene. I will begin by presenting a brief introduction on background modeling and a review of the different proposed methods. We conclude by discussing the applicability of these works to environmenta | |
| 2007.10.17 | Group seminar: Instrument Recognition in Monophonic and Polyphonic Music by Wanchi Lee (WPH 207 10/19/2007 1.00-2.00pm) |
|
Abstract: In this presentation, recently developed techniques in music instrument recognition as well as our improvement to the existing system will be introduced. Musical content analysis plays a key role in the musical information retrieval (MIR) system. In particular, musical instrument recognition provides useful information for automatic music transcription and music database indexing. Therefore, this problem has received a growing interest in recent years. Early researchers focused on the classification of isolated musical notes. The results of these early research fail to extend to real world problem since the audio features they used cannot be found in continuous musical instrument performance. Later, a more realistic application, which tries to identify the solo instrument played in continuous monophonic excerpt, draws more attention. Several recent research reported high correction rate on classification of some instruments, which is achieved by well-designed audio features and sophisticated feature selection techniques. More recently, instrument recognition in polyphonic has also started to be considered in some research, but the performance is still far from satisfactory. In the last part of the talk, our advance in this topic will be presented, which include using onset detection to assist the recognition. It will be shown that proper preprocessing and segmentation of music will facilitate the following classification tasks | |
| 2007.10.09 | Group seminar: Efficient Music Representation with Content Adaptive Dictionaries by Namgook Cho (WPH 207 10/12/2007 1.00-2.00pm) |
|
Abstract: An efficient music representation based on the matching pursuit (MP) technique with content adaptive dictionaries (CADs) will be discussed in the presentation. The Gabor atoms are commonly adopted in the MP-based signal representation due to their excellent time-frequency localization property. However, the Gabor dictionary may not yield a concise representation for music signals. Music signals have special characteristics as they are generated by musical instruments and/or human vocal tracks with musical rules and they are specified by pitches and durations of music notes. We exploit music characteristics to create content-adaptive dictionaries (CADs) that are tailored to instrument types and note pitch levels. As a result, we are able to project musical signals onto a subspace spanned by atoms from CAD for a concise and efficient representation. The proposed CAD representation technique is applied to music enhancement with noisy background to demonstrate the power of the proposed representation | |
| 2007.10.01 | Group seminar: Hierarchical Cloud Simulation based on Similarity Approach by Bei Wang (WPH 207 10/5/2007 1.00-2.00pm) |
|
Abstract: Cloud is attractive to peoples since it provides more realistic outdoor scenes. However, due to its amorphous shape, it is difficult to model and render in computer graphics field. For this kind of physical based problem, there are generally two kind of methods to do the modeling, one is the CFD which got to solve bunch of PDE equations, this method can generate astonishing result however computation cost is big. Another method is called procedural method, which is use noise function to generate the amorphous character, but their flexibility and extension is limited. In my talk, I am going to provided an efficient simulation method to achieve cloud simulation,especially the cumulus cloud simulation which is driven by convection. In atmospheric science, thermal is studied as a basic flow unit for cloud simulation, whose dynamic property can be described by similarity approach; however, this is not enough for computer graphics for visual details, thus inner motion of thermal will be introduced, Votex particle method for axis-symmetric flow will be presented | |
| 2007.09.25 | Group seminar: Realistic Fluid Simulation by Youngmin Kwak (WPH 207 9/28/2007 1.00-2.00pm) |
|
Abstract: Special effects (SFX) are used in many areas such as film, television, and game to realize realistic and dramatic scenes which are not possible by normal means. Fluid simulation is the core of SFX and includes liquid, smoke, and bubble simulations. Navier-Stokes equations (NSEs) are non-linear partial differential equations (PDEs) which describe the movement of fluids. The fast and visually accurate solution of those equations is the goal of fluid simulation in computer graphics area. Particle level set (PLS) method and advanced PLS will be introduced during the semina | |
| 2007.09.18 | Group seminar: Gesture recognition with Wii Remote by Pei-ying Chiang (WPH 207 9/21/2007 1.00-2.00pm) |
|
Abstract: A new revolution in game industry is coming in the wake of Wii's succeed. Previously, pursing high quality 3D animation was the main battleground in the gaming industry. However, the population of gamers has stagnated because most games were designed more and more complex. It would take a significant amount of time for a casual gamer to learn and memorize how to control and interact with the game. However, the innovation of Wii brings back the idea of "playing a game is simply for having fun". Moving your body exactly like what you want to do in the game, makes the game controls more straight forward and intuitive. In this presentation, I am going to talk about why Wii is so successful, what is the secret weapon of Wii? What are the weaknesses of Wii? How can we overcome these weaknesses to benefit future gaming applications | |
| 2007.09.18 | Congratulations to Szu-Wei (Wesley) Lee for passing his Qualifying exam! |
|
Congratulations to Szu-Wei (Wesley) Lee for passing his Qual on 9/11/2007 | |
| 2007.09.12 | Group seminar: Robust Multimedia Multicast with Network Coding by Gracia Wang (WPH 207 9/14/2007 1.00-2.00pm) |
|
Abstract: There are several challenges of multimedia multicast and streaming. One is packet loss which is mainly due to network traffic congestion. The second is the heterogeneous receivers. The third is dynamic construction of the multicast tree architecture. Previous research focuses on coding methods such as Application Layer FEC (AL-FEC) to provide redundant packet for erasure correction, and complicated algorithms to construct path-disjoint multicast trees to provide path redundancy. In this presentation, a robust and efficient joint Network Coding (NC) and AL-FEC architecture is proposed. Benefiting from network coding which can achieves optimal multicast rate, the proposed multicast architecture results in better erasure protection compared with AL-FEC. Moreover, it simplifies the multicast tree construction and maintenance. Simulation with unequal protected H.264/SVC bit-stream shows that joint NC/AL-FEC obtains better PSNR for heterogeneous receivers than AL-FEC in erasure network | |
| 2007.09.11 | Congratulations to Jiansong (Jason) Chen for passing his Qualifying exam! |
|
Congratulations to Jiansong (Jason) Chen for passing his Qual on 9/10/2007. His thesis proposal title is "Robust V-BLAST Receiver Design and Performance Analysis for MIMO Communication Systems". His thesis Guidance committee includes: Jay Kuo (Chair), Xiaoli Yu (Co-Chair), Charles Weber, Giuseppe Caire and Ming-Deh Huang (Outside Member) | |
| 2007.09.05 | Congratulations to Yu-Hao (Roger) Chang for passing his defense! |
|
Congratulations to Yu-Hao (Roger) Chang for passing his defense today (9/5/07). Roger's thesis title is "Preconditioning Techniques for Efficient Ultra-Wideband (UWB) Communication Systems." His PhD dissertation committee includes: Jay Kuo (Chairman), Xiaoli Yu and Wlodek Proskurowski (Outside Member) | |
| 2007.09.04 | Congratulations to Ilya Eckstein for passing his defense! |
|
Congratulations to Ilya Eckstein for passing his defense today (9/4/2007). His thesis title is "Correspondence-based Analysis of 3D Deformable Shapes: Methods and Applications". His thesis committee includes: Jay Kuo (Chairman), Mathieu Desbrun, Gerard Medioni and Richard Leahy (Outside Member) | |
| 2007.09.04 | Congratulations to Kelvin Chou for passing his defense! |
|
Congratulations to Kelvin Chou for passing his defense today (9/4/2007). His thesis title is "Techniques for Peer-to-Peer Content Distribution over Mobile Ad Hoc Networks". His thesis committee includes: Jay Kuo (Chairman), Konstantinos Psounis and XiaoJiang Chen (Outside Member) | |
| 2007.08.20 | Jay Kuo Receives 2007 Okawa Foundation Research Grant |
|
Jay Kuo is a recipient of a 2007 Research Grant from the Okawa Foundation for Information and Telecommunications. The Okawa Foundation was established in Japan in 1986 to provide funding for and give recognition to new studies in the information and telecommunications fields. The Okawa Research Grant will be formally presented at a ceremony in San Francisco in October 2007Related links: | |
| 2007.08.20 | Congratulations to Yu Shiu for his successful PhD thesis defense! |
|
Congratulations to Yu (Shiu) for passing his defense on 8/20! His thesis title is "Digital Signal Processing Techniques for Music Structure Analysis". His final dissertation committee members are: Jay Kuo (Chair), Shri Narayanan and Elaine Chew | |
| 2007.05.01 | Congratulations to Ronald for his passing the Qualifying exam |
|
Congratulations to Ronald for passing his Qual on 5/1 (Tuesday). The title of his thesis proposal is "Protocol Design and Performance Analysis for Opportunistic OFDM/OFDMA QoS Networks". His Guidance Committee includes: C.-C. Jay Kuo (Chair), Antonio Ortega, Michael Nealy, Zhen Zhang and Ramesh Govindan (Outside Member) | |
| 2007.04.23 | Congratulations to Usman for his passing the Qualifying exam |
|
Congratulations to Usman for his passing the Qual exam on 4/23/2007. His thesis proposal title is "Ultra-Wide Band Impulse Radio with Super-Orthogonal Turbo Codes". His committee members include: C.-C. Jay Kuo (Chair), Jerry M. Mendel, Charles L. Weber, Xiaoli Yu and Remigijus Mikulevicius (Outside member). He did an excellent presentation which has impressed the committee members a lot | |
| 2007.04.03 | Congratulations to Athena for her passing the Qualifying exam |
|
Congratulations to Athena for her passing the Qual exam (4/3/2007). The title of Athena's thesis proposal is "Design and analysis of high-performance cooperative relay networks". Her Guidance Committee Members include Professors Keith M. Chugg, Michael Neely, Xiaoli Yu, Fengzhu Sun and Jay Kuo (Chair) | |
| 2007.02.08 | Congratulations to Pei-Kai Liao for his successful PhD thesis defense! |
|
Congratulations to Pei-Kai Liao for his successful passing the PhD thesis defense on 2/8 (Thursday). The title of his PhD thesis is ˇ§Distributed Edge and Contour Line Detection for Environmental Monitoring with Wireless Sensor Networksˇ¨ and his dissertation committee members are Professors Antonio Ortega, Ramesh Govindan and Jay Kuo (Chair) | |
| 2007.01.24 | Congratulations to Layla for her passing the Qualifying exam |
|
Congratulations to Layla Tadjpour for her passing the Qual exam (1/24/2007). The title of Layla's thesis proposal is "Performance Improvement and Analysis of Multiaccess OFDM Transceivers in Mobile Environments". Her Guidance Committee Members include Professors Xiaoli Yu, Charles Weber, Keith Jenkins, Igor Kukavica and Jay Kuo (Chair) | |
| 2007.01.22 | Congratulations to Alex for his passing the Qualifying exam |
|
Congratulations to Alex Chia-Chun Hsu for his passing the Qual exam (1/22/2007). The title of Alex's thesis proposal is "Coexistence Mechanisms for Existing and Next Generation Unlicensed Band Protocols". His Guidance Committee Members include Professors Antonio Ortega, Michael J. Neely, Leana Golubchik, Ramesh Govindan and Jay Kuo (Chair) | |
| 2006.12.06 | Congratulations to Bei for her passing the Qualifying exam |
|
Congratulations to Bei for her passing the Qual exam today (12/6/2006). Her thesis proposal title is "Techniques for Efficient Cloud Modeling, Simulation and Rendering". Her Guidance Committee includes Jay Kuo (Chair), Richard Leahy, Keith Jenkins, Krishna Nayak and Karen Liu (Outside Member) | |
| 2006.11.28 | Congratulations to Jae-Joon Lee for his successful PhD thesis defense! |
|
Congratulations to Jae-Joon Lee for his successful passing the PhD thesis defense on 11/28 (Tuesday). Jae-Joonˇ¦s Thesis title is "Aging Analysis in Large-Scale Wireless Sensor Networksˇ¨ and his final thesis committee includes Professor Bhaskar Krishnamachari, Professor Gaurav Sukhatme and Jay Kuo (Chair) | |
| 2006.11.20 | Congratulations to Kelvin for his passing the Qualifying exam |
|
Congratulations to Kelvin for his passing the Qualifying exam on 11/20 (Monday). Kelvin\ˇ¦s thesis proposal title is "Peer-to-Peer Content Distribution Systems over Mobile Ad Hoc Networksˇ¨ and his Qual exam committee members include Kai Hwang, Antonio Ortega, Konstantinos Psounis, Xiaojiang Chen (Outside Member) and Jay Kuo (Chair) | |