Издательство Hindawi, 2007, -463 pp.After a long incubation in academia and in very specialized industrial environments,in the last ten to fifteen years research and development of image processingand computer vision applications have become mainstream industrial activities.Apart from the entertainment industry, where video games and special effects formovies are a billionaire business, in most production environments automatedvisual inspection tools have a relevant role in optimizing cost and quality of theproduction chain as well. However, such pervasiveness of image processing and computer vision applicationsin the real world does not mean that solutions to all possible problemsin those fields are available at all. Designing a computer application to whateverfield implies solving a number of problems, mostly deriving from the variabilitywhich typically characterizes instances of the same real-world problem.Wheneverthe description of a problem is dimensionally large, having one or more of itsattributes out of the normality range becomes almost inevitable. Real-world applicationstherefore usually have to deal with high-dimensional data, characterizedby a high degree of uncertainty. In response to this, real-world applications needto be complex enough to be able to deal with large datasets, while also being robustenough to deal with data variability. This is particularly true for image processingand computer vision applications. A rather wide range of well-established and well-explored image processingand computer vision tools is actually available, which provides effective solutionsto rather specific problems in limited domains, such as industrial inspection incontrolled environments. However, even for those problems, the design and tuningof image processing or computer vision systems is still a rather lengthy process,which goes through empirical trial-and-error stages, and whose effectivenessis mostly based on the skills and experience of the designer in the specific field ofapplication. The situation is made even worse by the number of parameters whichtypically need to be tuned to optimize the performance of a vision system. The techniques which are comprised under the term soft computing(namely, neural networks, genetic and evolutionary computation, fuzzy logic, andprobabilistic networks) provide effective tools which deal specifically with theaforementioned problems. In this book, we focus on genetic and evolutionarycomputation (GEC) and try to offer a comprehensive view of how the techniques itencompasses can solve some of the problems which have to be tackled in designingimage processing and computer vision applications to real-world problems. In the rest of this chapter, we will offer a brief overview of the contents ofthe book. First, we will provide a quick introduction to the main EC paradigms,in order to allow subsequent chapters to concentrate more specifically on the descriptionof each application and on the peculiarities of the approach they describerather than on the basic approaches. Then, we will illustrate how the book, whichdoes not necessarily require sequential reading, has been organized, to make iteasier for readers to navigate through it and to find the topics which are moreinteresting to them.Introduction Low-level vision Midlevel vision High-level vision
Чтобы скачать этот файл зарегистрируйтесь и/или войдите на сайт используя форму сверху.