Genetic and Evolutionary Computation for Image Processing and Analysis
Stefano Cagnoni, Évelyne Lutton, Gustavo Olague · 2007
1.1. What is this book about? After a long incubation in academia and in very specialized industrial environ-ments, in the last ten to fifteen years research and development of image processing and computer vision applications have become mainstream industrial activities. Apart from the entertainment industry, where video games and special effects for movies are a billionaire business, in most production environments automated visual inspection tools have a relevant role in optimizing cost and quality of the production chain as well. However, such pervasiveness of image processing and computer vision appli-cations in the real world does not mean that solutions to all possible problems in those fields are available at all. Designing a computer application to whatever field implies solving a number of problems, mostly deriving from the variability which typically characterizes instances of the same real-world problem. Whenever the description of a problem is dimensionally large, having one or more of its attributes out of the “normality ” range becomes almost inevitable. Real-world ap-plications therefore usually have to deal with high-dimensional data, characterized by a high degree of uncertainty. In response to this, real-world applications need to be complex enough to be able to deal with large datasets, while also being robust enough to deal with data variability. This is particularly true for image processing and computer vision applications. A rather wide range of well-established and well-explored image processing and computer vision tools is actually available, which provides effective solutions to rather specific problems in limited domains, such as industrial inspection in controlled environments. However, even for those problems, the design and tun-ing of image processing or computer vision systems is still a rather lengthy pro-cess, which goes through empirical trial-and-error stages, and whose effectiveness is mostly based on the skills and experience of the designer in the specific field of application. The situation is made even worse by the number of parameters which typically need to be tuned to optimize the performance of a vision system. 2 Genetic and Evolutionary