Abstract : This tutorial is aimed at providing a general introduction to the use of probability in the design of vision modiules and its use in computational vision systems. Emphasis is placed on basic concepts and quantitative use. This includes hypothesis testing, likelihood and Bayes theorem. The importance of understanding data quality and the evaluation of algorithm performance is addressed via the use of measurement error, error propagation, covariance estimation and Monte Carlo techniques. The limitations of established methods will be explained, where necessary, in order to provide an in depth understanding of the use of quantitative probability as a scientific tool. Taken together, these methods form a framework for the design of computer vision algorithms, leading to a better understanding of what constitutes appropriate solutions to vision analysis tasks. It is intended that by the end of the tutorial the student will have sufficient familiarity with the basic concepts that they will be able to begin to recognise genuine novelty in published work in this area, as apposed to re-invention of existing statistical methods. They will also understand the appropriate use of these techniques. This in turn should facilitate, not only higher standards in their own work, but better reviewing of conference and journal publications for the next generation of vision researchers.

Intended Audience: The tutorial is designed for an audience with intermediate level mathematical skills, while the coverage of probability concepts ranges from introductory to intermediate. The material is therefore suitable for students and researchers whishing to gain a better understanding of the scientific basis for algorithm design.

Neil Thacker has a background training in experimental physics followed by 20 years experience in the area of computer vision and image analysis. During this time he has worked in areas as diverse as; computer vision, neural networks and pattern recognition, medical image analysis, and VLSI hardware design. He has published in excess of 200 (journal and conference) papers. This work is now distributed from the TINA open source computer vision pages www.tina-vision.net in the form of a three volume thesis on Visual Intelligence.

Abstract : In calendar 2007 66 papers in the category 'Machine Vision and Pattern Recognition' were submitted to the journal and of these, just half were accepted. Don will describe how the review process is implemented, what makes a paper likely to be accepted (or rejected!), and what steps can be taken to ensure that a minimum of revision is required before acceptance.

Don Braggins has been an independent consultant in Machine Vision since 1983. He helped to found the UK Industrial Vision Association in 1992 and the European Machine Vision Association in 2003. In 1991 the grade of Fellow of SPIE was conferred on him, and in 2005 he was asked to take on the role of Associate Editor for Machine Vision and Pattern Recognition for SPIE's peer-reviewed journal Optical Engineering.

Abstract: In recent years extensive research efforts have been devoted to sports video content analysis and applications due to their wide viewer-ship and high commercial potentials. Technologies and prototypes have been developed to automatically or semi-automatically analyze sports video content, extract semantic events or highlights, intelligently adapt, enhance and personalize the content to meet users' preferences and network/device capabilities. Many applications have been developed and used in broadcasting video enhancement such as multi-camera based 3D virtual sports events, virtual ads insertion for sports video, and motion analysis systems for sports training, etc. The aim of this tutorial is to provide a brief overview of general video content analysis techniques and a comprehensive overview of the technical achievements in the research area of sports video analysis and applications. We first cover feature extraction methods, which include low-level feature extraction, midlevel representation creation and high-level semantics detection in sports videos. Next, we present the state of the art in sports video analysis from the following three aspects: structure analysis, event detection, content adaptation and enhancement. We also address the issues concerning test data preparations and performance evaluations for sports video analysis systems. Based on the current technologies used in sports video analysis and the demands from real-world applications, future promising directions and research challenges are discussed at the end of the tutorial.

Intended Audience: This tutorial is intended for educators, researchers, engineers, students and people interested in gaining an overall understanding of video content analysis and sports video analysis and applications. The audience is required to have basic understanding of image, video, audio and text media and preliminary knowledge of signal processing and machine learning.

Speaker Biographies

Dr. Changsheng Xu
Institute for Infocomm Research, Singapore

Prof. Qingming Huang
Chinese Academy of Sciences, China