Programming and software development activities generally take the center stage in digital image processing projects. Although it is essential to have efficient application-specific algorithms, in any machine vision system, the hardware reliability, ambient lighting, lens characteristics, and quality of the captured image dictate the level of repeatability and accuracy achievable in the measured data. These factors and the relatively sparse literature on imaging hardware underpin the scope of Chapters 2 through 8.
Despite the growth of image processing literature, the evolution of many fundamental preprocessing concepts remains obscure. Many algorithms have well-established mathematical roots, but using them without any reference to the underlying assumptions may constrain their application potential. Chapters 9 through 13 provide a unified account for generating the kernels and algorithms used in the image calibration and processing literature and embedded in virtually all machine vision software. As indicated earlier, these derivations permit the effective use of the associated software tools and enable users to develop application-specific preprocessing algorithms.
The influence of many optical parameters such as depth of field and aperture are well understood, but the cumulative effects of lens distortions and image digitization by photosite array and uneven illumination are difficult to assess prior to setting up a machine vision environment. By using an analytical framework for lens and sensor MTF assessment, camera orientation, and choice of illumination configuration, some of the uncertainties in image capture can be overcome. A summary of some key parameters for the integration of "front-end" hardware is given in Secs. 14.1 through 14.4.