Analisi di immagini mediante PR e SR

Righi M.

Pattern Recognition  Image Analysis  Image Segmentation 

Scope of this dissertation is Image Analysis, applied to images coming from the probe microscopes and sonar sub-water detection. In particular, the contribution of the origi- nal work has been the combination of methods of super-resolution methods of pattern- recognition. We can better understand the possible impact of the results obtained, by con- sidering the role that images and computer vision have in several different investigation and diagnostic techniques. Very often indeed, the resolution of the images is not optimal, due to the instrumental limits (for example, limited number of detectors for the acquisition of the photons in the magnetic resonance imaging), or due to acquisition method (as in the case of the probe microscopies), or to the noise of various kinds, present in all acquisition techniques. The methods for enhancing the resolution are commonly known as super-resolution methods. They are not trivial, since they require a priori knowledge of the object to super-resolve, not always available. In this thesis, we have tried to develop methods of super-resolution coupled with me- thods of pattern-recognition that offered a priori knowledge of the object to super-resolve in the form of a more faithful model of the real object. The innovative combination of pattern-recognition methods and super-resolution led to the development of a new al- gorithm, called PRIAR (Pattern Recognition Augmented Image Resolution), and to its implementation, the Tool PRIAR. The tool has been used with images of stem cells ac- quired with the Atomic Force Microscope (AFM), a kind of scanning probe microscope, and with images of sub-water identification by sonar. The obtained results are encouraging and we hope they can be fruitfully applied to various areas ranging from biomedical, materials science, to object oriented identification, and also to forensic science.