ABSTRACT
Image segmentation techniques are based on the division of an image or a set of images in multiple regions or pixel sets according to somedetermined characteristic of the image or any of its computational properties, such as the intensities of the pixels in the color map or image gradient. Therewith, it is possible to obtain a more or less precise definition of the border between these regions and therefore proceed to its isolation for a more simple and precise analysis of the region of interest. A glimpse on the state of the art of this matter makes clear that, among the several different segmentation techniques developed, it is not obvious to state which one provides the best results for a generic use (Lakare, 2000, Unter et al. 2008). Depending on the image type, the segmentation approach must be previously studied in order to take full advantage of the image’s specific characteristics. However, even taking in account the most recent segmentation techniques developed, the user’s experience is still a preponderate factor that clearly influences the obtained results. Therefore, the development of an algorithm that can perform a fully automatic segmentation without taking in account some previous information about the images to segment and therefore constraining its use to a very specific application is still a work in progress. In Medical Imaging, segmentation is of major interest because it allows the location and tracking of tumors or other pathologies as well as real time guidance for computer assisted orthopedic surgery or the acquisition and computational representation of an organ or tissue, providing information about its properties and morphology. The fully automation of this process will
save significant amount of time both in the identification of the pathology and in surgery preparation. In addition, the integration of such an algorithm in a commercial software package will allow less qualified users to perform such kind of analysis without having to fully comprehend the relying problems behind the graphic user interface.
