Introduction

Industrial image processing is gaining more and more importance as a testing methodology. One of the most challenging and complex problems of industrial image processing is surface inspection, which is the process aimed at detecting a defect on a surface. Often, the surface to be inspected is inhomogeneous and of high contrast. Brightness fluctuations on the surface are common. Still, all defects need to be detected irrespective of other problems and without identifying regular objects as defects. There are a number of image processing systems that are able to carry out surface inspection more or less successfully. However, the requirements of industry are growing so rapidly and on such a large scale that existing systems can no longer satisfy the demand. The reason for this is not the computing capacity of an image processing system but the methods used for the recognition of defects. This book will present an approach to this problem that allows the development of an algorithm suitable for the recognition of a surface defect. This algorithm has been implemented as C-library functions for Seelector by hema electronic GmbH (a digital signal processing image processing system) [1] and as plug-ins for NeuroCheck (a PC image processing system) [2] and has been successfully tested in several applications. This algorithm will be presented in this book and demonstrated with numerous examples.