CALCULATION SYSTEMS DESIGN

Author: MORALES SÁNCHEZ JUAN.
Year: 2004.
University: POLITÉCNICA DE CARTAGENA.
Place of defense: ESCUELA TÉCNICA SUPERIOR DE INGENIERÍA DE TELECOMUNICACIÓN.
Place of preparation: ESCUELA TÉCNICA SUPERIOR DE INGENIERÍA DE TELECOMUNICACIÓN.

Summary: SUMMARY: The estimate of movement between images is an increasingly important need in the digital image processing. This problem, also known as recording image, is to find the best match between two spatial images. The registration of image is used in varied fields of science and engineering, such as remote sensing and telemetry systems, medicine, mapping or vision. The aparíción of new technologies in image acquisition, especially in the medical field, demand more advanced techniques for recording image to automate tasks of medical diagnosis and treatment, in particular, motion estimation techniques not rigid and merging of images from different sources. In this Doctoral Thesis deals with the problem of estimating movement between images is not rigid, with the proposal of a novel method for recording image consisting of two well-defined stages: the first stage in establishing a correspondence between the two fuzzy space images , and the second is obtained from the previous result correspondence space explicit medíante a method of regularízación Tikhonov. Both aspects both in combination and individually represent a contribution to the art. In the first stage of processing, draws a connection space, which is characterized by convex parametric models. These models define the region spatial correlation of greater verisimilitude, which may represent different geometries, as points and curves. Unlike methods based on characteristics, the method developed not extract features from ímágenes or makes use of segmentation techniques, but used a strategy closer to the methods based on areas (though used in the broad sense) since correspondence obtained with a measure of similarity and difference based on the correlation. The second phase will receive final motion vectors, which relate explicitly both images. Using Tikhonov regularization is to assume that the correspondence space estimate is a map of a continuous and smooth. The regularization of ad hoc method involves an iterative process converges to a solution continuous, smooth and in accordance with the restrictions imposed by space parametric models. The adjustment mechanism is related to the theory of deformable models, allowing fix a priori conditions curvature or smoothness of the map of correspondence space. Since the parametric models used are convex, the final result corresponds to a minimum overall functional. The absence of minimum local, on the other hand, is not itself from other non - parametric methods or minimization of energy, like elastic, fluid registration or registration by diffusion. Additionally, it has carried out the extension of the comprehensive approach scenario trídímensional (3D), or log volumes. This extension covers the proposal for new parametric models (point, curve, surface and their combination) and projection techniques on these models. The method of recording images / volume proposed has been validated in a long series of scenarios: synthetic images with noise and high non-linear deformations, stereoscopic photographs, moving images of a temporary nature, from the distortion of objects captured, and medical imaging and computerized tomography. The results show a good behavior effects, such as concealment of objects, changes in lighting, noise or distortion premises. Moreover, the consistency of the final record is good even in cases where the bad condition of the problem is high. In 3D, has been registering the volume of medical data tomog 8 raffia co 43rd mputerizada, as a contribution to medical diagnosis reliable and planning radiation therapy. Incorporating measures similarity Statistical nature, such as information sharing, which allows the proposed method can be used in the registration of medical imaging multimode, and merge data from different sources. This represents the future of this main line Doctoral Thesis.

Author: IBAÑEZ GONZALEZ JACINTO JAVIER.
Year: 2005.
University: POLITÉCNICA DE VALENCIA.
Place of defense: Dep. Sistemas Informaticos y Computacion.
Place of preparation: Universidad Politécnica de Valencia.

Summary: This memory is part of the line of investigation High Performance Computing for the calculation functions Arrays and its Application to the resolution of Differential Equations. This field of research is having a boom in recent years, providing solutions to a wide range of issues such as the simulation of physical phenomena or troubleshooting of control. The thesis focuses primarily on the development of methods and the implementation of algorithms calculated Functions Arrays, and its application to the resolution of Ordinary Differential Equations (EDOs) and Differential Equations Matriciales of Riccati (EDMRs). The starting point for the work done has been the study of the state of art in each of these areas, and propose new methods and algorithms that, in some way, improve existing ones. There were further explored various fields of application of such approaches. One of them, ropagación ray luminous fiber glass fotónico and the ability to control their behavior ando means newspapers, is one of the most promising prospects in the search for alternatives to technology for semiconductors. In this regard, the High Performance Computing allows, through the simulation, the behavior of light in the media, supporting experimental theoretical issues that are currently being investigated. To obtain quality software has been used a paradigm of life cycle-type semiautomatic. In addition, there have been numerous tests that have identified, by hand, the optimal parameters of the algorithms implemented, and second, to analyze the accuracy and efficiency of code implemented.

Author: ALBARRACIN FERREIRA JAIME OCTAVIO.
Year: 2005.
University: OVIEDO.
Place of defense: EDIFICIO VALDES SALAS.
Place of preparation: DEPARTAMENTO DE INFORMÁTICA.

Summary: At a business model through the models, relational and E / R, or even objects, some models are surprisingly fragmented and amorphous, both in its structure and in its functionality, which do not correspond to the natural Organization. The natural structure of the Organization, consistent and very defined, it is dedicated in this thesis, both the Accounting and the paradigm of Objects. The functionality natural transactional and also consistent, it is deducted equally from both the accounting and the reengineering. From paradigm Systems conceived as improving the paradigm objects, obtained here is a new data model concept, systemic, which will be known as the Data Model Reorientado Object or MODRO. Such MODRO if it is able to reflect the natural structure of the Organization, projecting into corporate databases, thus redirected to objects. Therefore, it naturally encapsulates the functionality of the Organization cohesionándola through reengineering. Thus, integrating the real structures and functionality, MODRO leads to the integration of the Organization. Moreover, it emerges new rules of integrity additional to those already known, although exclusive him MODRO. It also provides a means for greater granularity for reducing responsibilities resistant be encapsulated, and to raise the level of abstraction of reusable components.