Robust Computer Vision: Theory and Applications

Preface Computer vision is the enterprise of automating and integrating a wide range of processes and representations used for vision perception. It includes many techniques that are useful by themselves, such as image processing (transforming, encoding, and transmitting images) and statistical pattern classification (statistical decision theory applied to general patterns, visual or otherwise). Moreover, it also includes techniques for geometric modeling and cognitive processing. The field of computer vision may be best understood by considering different types of applications. Many of these applications involve tasks that require either work in a hostile environment, a high rate of processing, access and use of large databases of information, or are tedious for people to perform. Computer vision systems are used in many and various types of environments-from manufacturing plants, to hospital surgical suits, and to the surface of Mars. For example , in manufacturing systems, computer vision is often used for quality control. In this application, the computer vision system scans manufactured items for defects and provides control signals to a robotic manipulator to remove defective parts automatically. Current examples of medical systems being developed include: systems to diagnose skin tumors automatically, systems to aid neurosurgeons during brain surgery, systems to perform clinical tests automatically, etc. The field of law enforcement and security is also an active area for computer vision system development with applications ranging from automatic identification of fingerprints to DNA analysis. In a standard approach, statistical techniques in computer vision applications must estimate accurate model parameters despite small-scale noise in the data, occasional large-scale measurement errors (outliers), and measurements from multiple populations in the same data set. Increasingly, robust estimation techniques from statistics are being used to solve these parameter estimation problems. Ideally, these techniques should effectively ignore the outliers when estimating the parameters of a single population. In our approach, we consider applications that involve similarity where the ground truth is provided. The goal is to find the probability density function which maximizes the VII VIII Preface similarity probability. Furthermore, we derive the corresponding metric from the probability density function by using the maximum likelihood paradigm and we use it in the experiments. The goal of this book is to describe and illuminate some fundamental principles of robust approaches. Consequently, the intention is to introduce basic concepts and techniques of a robust approach and to develop a foundation, which can be used in a wide variety of computer …