During the last decade, we have witnessed a considerable increase in the demand for image analysis. The reasons for this are many.
Firstly, there is the technological evolution of sensors: a great increase in the resolution of linear and matrix CCD cameras, the ever-increasing availability of color based features, low cost powerful software, practically unlimited storage space, etc.
In the industrial sector, we should also cite the need to automate production monitoring: defects detection, removal of products which do not conform to standard, controlling robots which allow us to replace defective parts, controlled handling and “packaging” of the products, all on an ever-growing scale.
In the medical sector, new three-dimensional imaging systems have emerged, which need assistance in the interpretation of the images and the management of enormous databases. Also, the domain of CAMIs (Computer Aided Medical Interventions) has been developed by using 3D navigation software.
In comparison to this ballooning demand, we can see that the evolution of image-analysis and processing software has not been as widespread, resulting in a number of disappointments for potential users.
There are many reasons — more fundamental than technological — for this slowness to develop: indeed, the vast majority of the images being analyzed are digitized, and they can be considered as defined functions on a two-dimensional (and sometimes three-dimensional) domain. However, ...