Research in "Representation, Analysis and Description of 2D and 3D Images" is done within the ICT project "Multimodal and Multidimensional Content and Media (ICT.P10)".
In particular, during 2007 the activity has focused mainly on methods for segmentation and representation, reaching significant goals. In the framework of watershed-based segmentation, efficient techniques have been suggested to reduce over-segmentation. This has been done in two different ways: by means of new criteria to evaluate region significance, or by resorting to multiscale image representation. Case-based-reasoning has also been used to improve the performance of segmentation algorithms and allow their use in a wider application domain. Finally, alternative approaches to segmentation, based on histogram analysis, have been suggested. In the framework of discrete representation schemes for 3D objects, 3x3x3 operators have been introduced for the computation of the Surface Skeleton. To this aim, the distance transform of the object has been used, which allowed us to obtain topologically correct results at a limited computational cost. Other significant results have been achieved in research fields closely related to the previous research activities. In particular, statistical features of grey-level images have been evaluated and used to measure the similarity degree of images when using case-based-reasoning, morphological-geometrical features have been computed in the framework of segmentation and geometrical-topological features for image representation. Finally a classification scheme has been proposed to classify voxels in complex 3D surfaces.