DIGITAL IMAGES

Author: LILLO SAAVEDRA MARIO FERNANDO.
Year: 2004.
University: POLITÉCNICA DE MADRID [www.upm.es].
Place of defense: LECTURA DE INFORMATICA.
Place of preparation: FACULTAD DE INFORMATICA.

Summary: There is a wide range of applications in the fields of remote sensing, which requires the availability of images with high spatial and spectral resolution. One way to dispose of such products, costs relatively accessible, it ia use of fusion imaging techniques. Obviously, that to obtain accurate results of a proper, it is essential to define in an objective manner, the quality of the resulting images. In literature, are described a wide range of fusion techniques of images Even if the theoretical underpinning in which most of these techniques would the merger of many different types of satellite imagery, a very high percentage of applications restricted to the merger of a multispectral image with a panchromatic, in this Doctoral Thesis, are designed, evaluated and implemented new strategies for merging panchromatic and multispectral imagery in a way that integrates objective information from the source images through establishment of a compromise between the spatial and spectral quality of the final image merged. In the first phase of this thesis, raised new strategies for merging images based on the Model Mixed Linear (MML). That certain functions defined goals which enabled control the contribution of information from the source image to the image merged, and were optimized using stochastic algorithms, such as the Simulating Annealing and Genetic Algorithms, in a second stage, was raised a new strategy for merging using the Fourier transform using filters tailored to allow control the quantity and quality of information that should be incorporated into the merged image. Two criteria were proposed targets for tuning the filters, based on levels of entropía and power of the images source. In addition, it raised another merger strategy based on wavelet transform, calculated using the algorithm Mallat. This adjusted coefficients associated with the mother wavelet characteristics of the source images. These coefficients were estimated from the parameters associated with type FIR filters, estimated in the merger by the Fourier transform. To estimate the spectral quality of the images merged a quantitative manner, it uses the index ERGAS, while for the evaluation of quality space, proposed a new index, based on the above and as defined in the same domain as the ERGAS spectral This has resulted in a compromise between the spatial and spectral quality of the images merged. Moreover, a new strategy was proposed merger based on wavelet transform, calculated using the algorithm á trous. Based on the indices of quality ERGAS, either version as spectral space, identified weights of the wavelet coefficients associated with the panchromatic image, in order to establish a mechanism that allows control the compromise between the spectral and spatial information, images produced by the source. In a third phase were carried out qualitative and quantitative comparisons between the images by merging the strategies of merger proposals and images obtained from the merger strategy based on the trans 8 formed 63e of Wavelet Mallat standard. This analysis allowed us to have a thorough knowledge of the strengths and weaknesses of various strategies studied, as well as the optimal use of each of them, which allowed select the most suitable for a particular application. In the final phase of this thesis, addressed a specific problem: determining the level of density of vegetation in vineyards using satellite imagery and original merged. This application was to experiment in real problems, the goodness of the images merged with the selected strategy. The results throughout this thesis, it can be concluded that mergers obtained through methodologies and schemes proposed, in a balanced manner integrate information from the source images, thus ensuring an image with a compromise between his objective quality space and spectral.

Author: DOSIL LAGO RAQUEL.
Year: 2004.
University: SANTIAGO DE COMPOSTELA [www.usc.es].
Place of defense: ESCOLA TÉCNICA SUPERIOR DE ENXEÑERÍA.
Place of preparation: DEP. DE ELECTRÓNICA E COMPUTACIÓN. FACULTAD DE FÍSICA.

Summary: The work presented in this thesis is an attempt to fill the space between the multiple detectors simple features that exist for volumetric data and processing techniques at a high level. The method developed offers both a model of representation as a technique for the detection of low-level features, which we call visual patterns. The visual patterns, defined here as the composition of elemental characteristics of frequency and orientation, provide a more complete description of the characteristics of an image produced by the detectors features simple. For detection, we propose the use of a strategy descomposición-integración general purpose and entirely driven by data, ie, not using a priori information. This implies, first, the development of a model for the decomposition of a 3D image in a set of basic features, which should produce a comprehensive representation of the image. To this end, we adopted a multirresolución scheme, in which the basic characteristics, which we call frequency characteristics are sintotizadas at different scales and directions. For its part, the integration stage is to identify what are the characteristics often associated with each pattern visual and reintegrated, thereby generating a representation separate from those patterns. This stage is raised as a problem of unsupervised clustering, or grouping, based on a measure of distance between frequency characteristics reflecting the approach of integrating features. Several authors argue that the human visual system integration feature is based on the Phase Congruence offered. This is the approach used in this work.

Author: GUTIÉRREZ AGUADO JUAN.
Year: 2005.
University: VALENCIA [www.uv.es].
Place of defense: ESCOLA TÈCNICA SUPERIOR D'ENGINYERIA.
Place of preparation: ESCOLA TÈCNICA SUPERIOR D'ENGINYERIA.

Summary: CROWD OF TECHNIQUES FOR THE SOLUTION OF PROBLEMS IN TREATMENT OF IMAGES REQUIRING A MODEL OF THE STATISTICAL SIGNAL TO PROCESS. IN THIS THESIS, ANALIZAMOS THE RELATIONSHIP BETWEEN THE EXISTING OPERATION OF CORTEX VISUAL AND STATISTICS OF THE IMAGES NATURAL AND PROPOSE THE USE GENÉRICA OF MODELS OF RESPONSE CORTICAL AS ALTERNATIVE TO THE USE OF MODELS IN THE STATISTICAL APPLICATIONS CODING OF VIDEO (FOR DESIGN OF CUANTIZADOR AND IN ESTIMATING THE MOVEMENT) AND RESTORATION OF IMAGES DEGRADED (TO DEFINE CRIMINALIZATION OPERATORS IN THE CONTEXT OF ADJUSTMENT).

Author: NASSABAY PARDO SALUA ESTHER.
Year: 2006.
University: GRANADA [www.ugr.es].
Place of defense: E.T.S. DE ING.INFOR. Y DE TELECO..
Place of preparation: ESCUELA TÉCNICA SUPERIOR DE INGENIERÍAS INFORMÁTICA Y DE TELECOMUNICACIÓN.

Summary: Summary At present, the signal processing, including image processing becomes increasingly important. There are many areas where improvement is needed visual information in order to obtain good results. Of these fields may be highlighted branches of the medical X-ray images, tomagrafías, fMRI, etc., biology, geography, archeology, astronomy, defense and military forces, and so on. In general the noise level is presented as a mixture of random signals with different frequencies which may be higher or lower than the signal of interest. The problem of separation blind signal is the recovery of the original signals from mixtures detected by sensors, knowing only the latter. This paper aims at doctoral and its importance lies in the development of a comparative study of different algorithms used, based on the technical analysis of separate components applied to images with noise. This work will be conducted to develop a sequence which specifies the field of classification (PDI), the level of the problem that allows a focus on the elimination of noise in images; level approaches to the solution, which shows the various methods available for addressing the problem of noise, and finally the level of performance that seeks to provide targeted solutions. This report is organized as follows: Chapter 1: Introduction to the analysis of separate components and the separation blind signal. This chapter presents the definition and formulation of mathematical technique Independent Component Analysis applied to the problem of Blind Signal Separation. First there are the statistical basis and its importance in the separation of signals. Then develops information theory and the basics needed for the technical ICA. He then introduces the techniques of Principal Component Analysis (PCA) and the Independent Component Analysis (ICA) with its corresponding mathematical definitions, restrictions and algorithms. Finally a comparison is made between ICA and PCA. Chapter 2: Connection between the human visual system (SVH), the Independent component analysis (ICA). Is this chapter provides the connection between ICA and image processing. In the first instance is a summary history retaking the last years of research in image processing with an introduction to the human visual system. After introducing the ICA connection with the study of the images. Chapter 3: Algorithms ICA. This chapter develop algorithms ICA most commonly used technique by different investigators, explaining their advantages and disadvantages in order to pave the way for its use in the next chapter. Chapter 4: The noise at ICA. This chapter assesses the connection between ICA and noise, he examines the behavior of five different algorithms ICA (fastICA, JADE, AMUSE, fastGEO and dss_fastICA), in response to three different kinds of noise (gaussian, multiplicative and impulsive (salt and pepper)), applied to the area of digital image processing. In particular is studied as they behave different filters matrices A mixture obtained from the observations and influence of each type of noise on the signal sample. An assessment is also made of sensitivity that allows end in a catalogalización of different tools. Chapter 5: Applications: Automatic detection filter noise and De-noising. This chapter was developed for detection automatic classification of filters, representing both interesting information as information only noise 8. Igualm 36th body shows application of eliminating noise. In the "Conclusions and Future Work" summarizes the major contributions that have been made in this area with the development of this doctoral work.

Author: NASSABAY PARDO SALUA ESTHER.
Year: 2006.
University: GRANADA [www.ugr.es].
Place of defense: E.T.S. INGENIERÍAS INFORMÁTICA Y DE TELECOMUNICACIÓN.
Place of preparation: ESCUELA TÉCNICA SUPERIOR DE INGENIERÍAS INFORMÁTICA Y DE TELECOMUNICACIÓN.

Summary: At present, the signal processing, including image processing becomes increasingly important. There are many areas where improvement is needed dela visual information in order to obtain good results. Of these fields may be highlighted branches of the medical X-ray images, tomagrafías, fMRI, etc., biology, geography, archeology, astronomy, defense and military, ect. In general the noise level is presented as a mixture of random signals with different frequencies which may be higher or lower than the signal of interest. The problem of separation blind signal is the recovery of the original signals from mixtures detected by sensors, knowing only the latter. This paper aims at doctoral and its importance lies in the development of a comparative study of genetic algorithms used different based on the technical analysis of separate components applied to images with noise. This work will be conducted to develop a sequence which specifies the field of classification (PDI), the level of the problem that allows a focus on the elimination of noise in images; level approaches to the solution, which shows the various methods available for addressing the problem of noise, and finally the level of performance that seeks orientares to provide solutions. This report is structured as follows: CHAPTER 1 Introduction to the analysis of separate components and the separation blind signal. This chapter presents the definition and formulation of mathematical technique Independent Component Analysis applied to the problem of Blind Signal Separation. First there are the statistical basis and its importance in the separation of signals. Then develops information theory and the basics needed for the technical ICA. He then introduces the techniques of Principal Component Analysis (PCA) and the Independent Component Analysis (ICA) with its corresponding mathematical definitions, restrictions and algorithms. Finally a comparison is made between ICA and PCA. CHAPTER 2 connection between the human visual system (SVH), the Independent component analysis (ICA). This chapter presents the connection between ICA and image processing. In the first instance is a summary history retaking the last years of research in image processing with an introduction to the human visual system. After introducing the ICA connection with the study of the images. CHAPTER 3 Algorithms ICA. This chapter develop algorithms ICA most commonly used technique by different investigators, explaining their advantages and disadvantages in order to pave the way for its use in the next chapter. CHAPTER 4 Noise in ICA. This chapter assesses the connection between ICA and noise, he examines the behavior of five different algorithms ICA (fastICA, JADE, AMUSE, fastGEO and dss_fastICA), in response to three different kinds of noise (gaussian, multiplicative and impulsive (salt and pepper)), applied to the area of digital image processing. In particular is studied as they behave different filters matrices maxla In obtained from the observations and influence of each type of noise on the signal sample. An assessment is also made of sensitivity that allows end in a catalogalización of different tools. CHAPTER 5 Applications: Automatic detection filters as noise and De-noising. This chapter was developed for detection automatic classification of filters, both representing information 8 interest 2a2 before as information only noise. Also shown implementation of eliminating noise. In the section 'Conclusions and future work' summarizes the main contributions that have been made in this area with the development of this doctoral work.

Author: VARGAS VÁZQUEZ DAMIÁN.
Year: 2006.
University: POLITÉCNICA DE MADRID [www.upm.es].
Place of defense: FACULTAD DE INFORMÁTICA.
Place of preparation: FACULTAD DE INFORMÁTICA.

Summary: The processing and analysis of images is a field in constant innovation. One of its goals is to improve the quality of the images to make it easier to find information. Once modified and improved the image, you begin a process that allows segmentation break it into regions of interest for classification. The segmentation of images is grouping regions in visual terms known and its result is used by multiple applications of all kinds. The work of this report focuses on the domain of application in the field of biomedical imaging. Target and accurately reconstruct the structures of this type of images is difficult because of the complexity and variability of the anatomical structures and cellular forms of interest. The Morphology Mathematics (GM) has a special place in the analysis and image processing. This work experience with a series of transformations and tools based on GM For the segmentation of images. On the one hand, the proposed amendment and subsequent use of a new class of morphological filters: emph (openings and closures judiciously reconstruction). Such filters include an additional criterion which allows you to control the reconstruction process. In particular, it may limit the excessive reconstruction of flat areas that occurs occasionally in the changes for reconstruction. Moreover, it is possible to separate, if you like, certain regions of the image. Using techniques GM To segment an input image, are taking advantage of the properties referred to calculate markers significant regions. These properties are used for the segmentation of medical images, but can be used in other fields. On the other hand, proposes a new method for segmentation based on transformed "watershed". This approach implements a process of reconstructing images from labeled markers, the gradient of the image is regarded as a topographic relief that is flooded (similarly in the case of the "watershed" Traditional). However, it incorporates a criterion in the reconstruction process by allowing additional flexibility in the separation of structures of interest. Presents also has two additional methods that provide new solutions and applications thanks to the modification of certain parameters of the original.

Author: LOPEZ CORONADO OCTAVI.
Year: 2006.
University: AUTÓNOMA DE BARCELONA [www.uab.es].
Place of defense: Facultad de Ciencias.
Place of preparation: UNIVERSITAT AUTÃ’NOMA DE BARCELONA.

Summary: In this work we have investigated various techniques for generating optical elements difractivos that change the profile of intensity along the axis and the properties of apodización and hiperresolución cross an optical system, as well as its implementation through a spatial light modulator liquid crystal (LCSLM). First are designed and implemented complex transmission pupils not uniform, with the aim of changing the profile of the intensity in the vicinity of the plane image of a convergent system and achieve arbitrary profiles with features of interest, such as rectangular different profiles that increase DOF or providing an answer bifocal the optical system. The pupils have been implemented in a spatial light modulator in configuration only through a phase encoding method under complex functions. It has been experimentally evaluated numerically and the results of these pupils by comparing the intensity profile with the desired profile obtained along the axis, and has analyzed the responses to the image of a coherent point with lighting at various levels cross. It has been experimentally verified that the technique is especially useful to stop the increase in the depth of focus of the system. In the second part of this work have proposed different methods for implementing different focal lenses difractivas of in a single LCSLM in fashioning only phase modulation, and assessed their response in relation to the increase in the DOF and behavior hiperresolvente or apodizante at the cross. Experimental results have been obtained for both the image of a coherent point with lighting as images of objects with extensive lighting inconsistent. It has been observed that, for mounting optical used, the method of random multiplexing that combines 33 lens focal consecutive is the most satisfactory for increasing DOF of the optical system. This option provides a multiplexed profile uniform intensity over the entire DOF and structure of the transverse image of a point (PSF) that is uniform throughout this interval, so it is very suitable for training extensive imaging.