Many computer vision and multimedia applications such as immersive multimedia systems, free viewpoint video or TV, 3D reconstruction and video surveillance employ multiple cameras for acquiring information of scenes. In most cases, cameras are arranged based on experience or in an ad-hoc way, for instance, a regular grid of cameras for free-viewpoint video and parallel stereo cameras for depth estimation. While the geometry of a multi-camera system has been well studied, little research has been reported on how well a multi-camera system fits to the scene. In this talk, two newly developed concepts, effective sampling density (ESD) and correspondence field (CF), will be introduced. ESD is a tractable metric that jointly quantifies how well the cameras sample the scene and the acquired samples are used in the processing (e.g. rendering in free-viewpoint video) and CF is a new mathematical representation of camera geometry that can be used to quantify the suitability of a camera configuration for a given arrangement of objects in the scene. We will demonstrate that ESD can be used to estimate signal distortion for a grid-camera based free viewpoint video system and, hence, the output video quality without access to the ground truth or actual implementation of the system and that CF can be used to optimize camera configuration to improve depth estimation.
Wanqing Li received his PhD in electronic engineering from The University of Western Australia. He was a Principal Researcher at Motorola Lab from 1998-2003 and a visiting researcher at Microsoft Research, Redmond in 2008, 2010 and 2013. He is currently an Associate Professor and Co-Director of Advanced Multimedia Research Lab (AMRL) of University of Wollongong, Australia. His research areas include 3D computer vision, 3D multimedia signal processing and medical image analysis.
Dr Li is a Senior Member of IEEE and currently a co-chair of the 3D Rendering, Processing and Communications Interest Group, Multimedia Technical Committee of IEEE Communication Society. He is the guest editor of the special issue on Human activity understanding from 2D and 3D data (2015), International Journal of Computer Vision, and the special issue on Visual Understanding and Applications with RGB-D Cameras (2013), Journal of Visual Communication and Image Representation. He served as a Co-organizer of the IEEE International workshop on Human Activity Understanding from 3D Data (HAU3D) (2011-2013) and Hot Topics in 3D multimedia (Hot3D) (2014), an area chair of International Conference on Multimedia & Expo (ICME) 2014, a publication chair of IEEE Workshop on Multimedia Signal Processing (MMSP) 2008, General Co-Chair of ASIACCS'09 and DRMTICS'05, and technical committee members of numerous international conferences and workshops including CVPR, ICME, ICIP, MMSP and 3DTV-Con.