Keynote Speakers at PSIVT 2011

Abstract:

Television realized a human dream of seeing a distant world without actually going there. However, it allows the users to see only a single view of a 3D scene they want to see. Furthermore, the viewpoint can’t be changed by the users. Stereoscopic 3DTV also has the same limitation although it transmits 2 views for right and left eyes and gives us 3D sensation. Free-viewpoint Television (FTV) is an innovative visual media that breaks down this limitation of the current visual media and allows the users to see a 3D scene by freely changing the viewpoint. FTV is the ultimate 3DTV and ranked as the top of visual media since it has infinite number of viewpoints.

We proposed the concept of FTV and constructed the world’s first real-time FTV system including the complete chain of operation from image capture to display. FTV captures rays sparsely by using multi-cameras set discretely in the 3D space and generates the remaining rays. For this purpose, we have developed new video technologies such as technology that treats many cameras as if they are a single camera, ray integration and interpolation technologies and so on. We have also developed an all-around dense ray capture method and an efficient ray capture method that captures rays efficiently with reduced number of pixel data. At present, FTV is available on a single PC or a mobile player. All-around ray-reproducing 3DTV and FTV with free listening-point audio are also realized.

We proposed FTV to MPEG in 2001. MPEG regarded FTV as the most challenging 3D media and started the international standardization activities of FTV in 2004. The first phase of FTV was Multi-view Video Coding (MVC) and the second phase of FTV is 3D Video (3DV). MVC enables the efficient coding of multiple camera views and was completed in 2009. MVC has been adopted by Blu-ray 3D. 3DV is a standard that targets serving a variety of 3D displays. “Call for Proposals on 3D Video Coding Technology” was issued in March 2011.

Biography:

Masayuki Tanimoto received his B.E., M.E., and Dr.E. degrees in Electronic Engineering from the University of Tokyo in 1970, 1972, and 1976, respectively. He joined Nagoya University in 1976. Since 1991, he has been a Professor at Graduate School of Engineering, Nagoya University. He has been engaged in the research of image coding, image processing, 3D imaging, FTV and ITS.

He was the president of the Institute of Image Information and Television Engineers (ITE), and a fellow of the Institute of Electronics, Information, and Communication Engineers (IEICE) and ITE. He received the ITE Distinguished Achievement and Contributions Award, the IEICE Achievement Award, and the Commendation for Science and Technology by the Minister of Education, Culture, Sports, Science, and Technology.

Abstract:

For the 3D reconstruction of static and dynamic scenes, we have developed several robust vision methods ranging from low-level edge detection to the design of novel camera systems.

In this talk, we first present a physics based edge detection and a dense stereo matching method based on the characteristics of the human vision system. Specifically, a novel image noise model based on the Skellam distribution is very effective to the edge detection problem.

The second part of the talk introduces new camera systems to capture the 3D information accurately and efficiently. The camera systems include (1) a bi-prism stereo camera, (2) “camera + depth” fusion camera systems, (3) a fast bundle adjustment based approach for large-scale datasets, (4) a novel coded light for dynamic scene. Specifically, we present a hand-held fusion sensor system, consisting of four cameras and two 2D laser scanners, to capture 3D information of large-scale scenes. This new approach allows boosting the advantages of two sensor systems and complements the weakness of the two.

As an important application of 3D vision techniques, we demonstrate the robustness of the methods by automatically synthesizing high-quality novel stereoscopic views from video and/or 3D information.

Biography:

Prof. InSo Kweon received his B.S. and M.S. degrees in Mechanical Design and Production Engineering from Seoul National University, Seoul, Korea, in 1981 and 1983, respectively, and the M.S. and Ph.D. degree in Robotics from the Robotics Institute at Carnegie Mellon University, Pittsburgh, U.S.A, in 1986 and 1990, respectively. He worked for Toshiba R&D Center, Japan, and joined the Department of Automation and Design Engineering at KAIST in 1992. He is now a Professor in the Department of Electrical Engineering at KAIST. His research interests are in computer vision, robotics, pattern recognition, and automation. Specific research topics include invariant based visions for recognition and assembly, 3D sensors and range data analysis, color modeling and analysis, robust edge detection, and moving object segmentation and tracking. He is a member of ICASE, IEEE, and ACM.

Abstract:

In this presentation, we aim to provide an overview of the mathematical theory of Compressed Sensing (CS), an emerging field in Signal Processing and Information Theory. CS is deemed to have provided a new signal acquisition framework with which the samples of a given signal of interest can be taken while being compressed simultaneously and the given signal can be regenerated perfectly with much smaller number of measured samples than what would be needed in the classical sampling approach. Thanks to this new capability, the idea of CS has been applied and shown successful in quite a number of engineering problems, where sample taking is expensive or time consuming, and where the number of sensors is limited perhaps due to space limitation, including but not limited to wideband radars, ultra wideband spectrum sensing, functional MRIs, and Terahertz imaging. As such, understanding the mathematics of the CS theory becomes quite important and interesting. In CS, sample taking is done via linear projection of a given signal against a prescribed set of kernels, i.e., one linearly projected sample per kernel; the canonical approach to recover the original signal from the projected samples in CS is to use the so-called Basis Pursuit algorithm, a linear programming based L1 minimization algorithm. Today, we have many kernels and recovery algorithms available. It is not difficult to see that the performance of a CS system depends heavily on the choice of the kernel and the recovery algorithm. Our overview will include discussion of kernels and recovery algorithms, and certain theoretical tools with which one can determine the number of measurements needed for perfect recovery given a kernel.

Biography:

Prof. Heung-No Lee graduated from UCLA obtaining his Ph.D. degree in Electrical Engineering in 1999. He received B.S. and M.S. degrees in Electrical Engineering at UCLA as well, in 1993 and 1994, respectively. He then moved to HRL Laboratory, Malibu, California, and worked there as a Research Staff Member from 1999 to 2002. He was then appointed as Assistant Professor at the University of Pittsburgh, Pittsburgh, Pennsylvania, in 2002 where he stayed till 2008. He founded Communications Research Lab in the Electrical and Computer Engineering Department. He obtained funding from various sources, including National Science Foundation, the Technology Collaborative, the US Army, and private companies. Three Ph.D. students and four M.S. students have graduated from his research programs at the University of Pittsburgh. He then moved to Gwangju Institute of Science and Technology (GIST) in Jan. 2009. The general areas of his research lie in Signal Processing Theory, Communications, and Information Theory, and their application to Wireless Communications and Networking, Compressive Sensing, Future Internet, and Brain Computer Interface.