In this study a fully automatic and scale invariant feature extractor which does not require manual initialization or special equipment is proposed. Face location and size is extracted using skin segmentation and ellipse fitting. Extracted face region is scaled to a predefined size, later upper and lower facial templates are used for feature extraction. Template localization and template parameter calculations are carried out using Principal Component Analysis. Changes in facial feature coordinates between analyzed image and neutral expression image are used for expression classification. Performances of different classifiers are evaluated. Performance of proposed feature extractor is also tested on sample video sequences. Facial features are extracted in the first frame and KLT tracker is used for tracking the extracted features. Lost features are detected using face geometry rules and they are relocated using feature extractor. As an alternative to feature based technique an available holistic method which analyses face without partitioning is implemented. Face images are filtered using Gabor filters tuned to different scales and orientations. Filtered images are combined to form Gabor jets. Dimensionality of Gabor jets is decreased using Principal Component Analysis. Performances of different classifiers on low dimensional Gabor jets are compared. Feature based and holistic classifier performances are compared using JAFFE and AF facial expression databases.

This thesis analyzes the human action recognition problem. Human actions are modeled as a time evolving temporal texture. Gabor filters, which are proved to be a robust 2D texture representation tool by detecting spatial points with high variation, is extended to 3D domain to capture motion texture features. A well known filtering algorithm and a recent unsupervised clustering algorithm, the Genetic Chromodynamics, are combined to select salient spatio-temporal features of the temporal texture and to segment the activity sequence into temporal texture primitives. Each activity sequence is represented as a composition of temporal texture primitives with its salient spatio-temporal features, which are also the symbols of our codebook. To overcome temporal variation between different performances of the same action, a Profile Hidden Markov Model is applied with Viterbi Path Counting (ensemble training). Not only parameters and structure but also codebook is learned during training.

Nano scale processing of IC chips has become the prime production technique as the microelectronic industry aims towards scaling down product dimensions while increasing accuracy and performance. Accurate control of temperature and a good monitoring mechanism for thickness of the deposition layers during epitaxial growth are critical parameters influencing a good yield. The two-fold objective of this thesis is to establish the feasibility of an alternative to the current pyrometric and ellipsometric techniques to simultaneously measure temperature and thickness during wafer processing. TAP-NDE is a non-contact, non-invasive, laser-based ultrasound technique that is employed in this study to contemporarily profile the thermal and spatial characteristics of the wafer. The Gabor wavelet transform allows the wave dispersion to be unraveled and the group velocity of individual frequency components to be extracted from the experimentally acquired time waveform. The thesis illustrates the formulation of a theoretical model that is used to identify the frequencies sensitive to temperature and thickness changes. The group velocity of the corresponding frequency components is determined and their corresponding changes with respect to temperature for different thickness are analytically modeled. TAP-NDE is then used to perform an experimental analysis on Silicon wafers of different thickness to determine the maximum possible resolution of TAP-NDE towards temperature sensitivity, and to demonstrate the ability to differentiate between wafers of different deposition layer thickness at temperatures up to 600?C. Temperature resolution is demonstrated for ?10?C resolution and for ?5?C resolution; while thickness differentiation is carried out with wafers carrying 4000? and 8000? of aluminum deposition layer. The experimental group velocities of a set of selected frequency components extracted using the Gabor Wavelet time-frequency analysis as compared to their corresponding theoretical group velocities show satisfactory agreement. As a result of this work, it is seen that TAP-NDE is a suitable tool to identify and characterize thickness and temperature changes simultaneously during thermal annealing that can replace the current need for separate characterization of these two important parameters in semiconductor manufacturing.

This thesis consists of two parts. In the first chapter, we give some results on modulation spaces. First the relationship between the classical spaces and the modulation spaces is established. It is proved that certain modulation spaces defined on R² lie in the BMO space. Another result is that the Zak transform, a discrete time-frequency transform, maps a modulation space into a higher dimensional modulation space. And by using these results, an uncertainty principle for Gabor frames via modulation spaces is obtained. In the second part, we deal with optimization of an increasing positively homogeneous functions on the unit simplex. The class of increasing positively homogeneous functions is one of the function classes obtained via min-type functions in the context of abstract convexity. The cutting angle method is used for the minimization of this type functions. The most important step of this method is the minimization of a function which is the maximum of a number of min-type functions on the unit simplex. We propose a numerical algorithm for the minimization of such functions on the unit simplex and we mathematically prove that this algorithm finds the exact solution of the minimization problem. Some experiments have been carried out and the results of the experiments have been presented.

Στην παρούσα εργασία γίνεται μια προσπάθεια οπτικής ανάκτησης λογοτύπων με σκοπό την οργάνωσή τους για μετέπειτα χρήση τους σε περιπτώσεις αναζήτησης ομοίων λογοτύπων από χρήστες. Η μελέτη γίνεται με χρήση μεθόδων και εργαλείων τα οποία είτε έχουν σχεδιαστεί για αυτό το σκοπό είτε έχουν ληφθεί από άλλα κομμάτια του τομέα αναγνώρισης προτύπων. Πιο συγκεκριμένα χρησιμοποιούνται οι μεθοδολογίες: α. Angular Radial Transform (ART) που ορίζεται στο MPEG-7 πρωτόκολλο. β. Ένας αλγόριθμος σύγκρισης σχημάτων ως κατανομές ιστογραμμάτων που τον ονομάζουμε Angular Radial Histogram (ARH) και περιγράφεται σε δημοσίευση που αφορά την αναγνώριση σχημάτων. γ. Μια μεθοδολογία που εκμεταλλεύεται τον αλγόριθμο DCT και την συμπιεσμένη φύση των αρχείων εικόνων - λογοτύπων στο σημερινό web για γρήγορη αναζήτηση η οποία έχει χρησιμοποιηθεί για οργάνωση βάσεων γενικών εικόνων. δ. Μια μεθοδολογία στηριγμένη στα φίλτρα Gabor και το WW-Test η οποία έχει εφαρμοστεί με αρκετά καλά αποτελέσματα σε πεδιά αναγνώρισης προτύπων. Η έρευνα γίνεται σε τρεις (νεοκατασκευασμένες) βάσεις λογοτύπων όπου τα αποτελέσματα φαίνεται να είναι θετικά ειδικά για την τελευταία μεθοδολογία. / In this work, four different algorithms are used in the field of logos retrieval, in an efford to organize logos databases that can be handful in cases where users may need to retrieve similar logos. The algorithms that are used are a. Angular Radial Transform (ART) which is defiend in MPEG-7, b. An algorithm that compairs logos outlines called sas Angular Radial Histogram which has been proposed for general shape matching procedures, c. An algorithm based on the well known DCT and the compression nature of the very common JPEG protocol, d. An algorithm based on Gabor filters for feature extraction and the Wald-Wolfowitz for feature comparisons, which gave very good results. The tests took place on three different logos database which were constructed for this purpose.

Dans certains contextes, les méthodes classiques utilisées pour le calcul de champs rayonnés ou diffractés en présence d'obstacles de grande taille par rapport à la longueur d'onde, comme l'Optique Physique ou les méthodes de rayons, ne sont pas valides ou deviennent très lourdes en temps de calcul. La théorie des frames de Gabor fournit un cadre rigoureux permettant de décomposer une distribution de sources électromagnétiques, définie dans une ouverture équivalente plane, en une somme plus ou moins redondante de fenêtres gaussiennes. Cette décomposition peut servir de base à des algorithme de lancer de faisceaux gaussiens.Jusqu'à présent cette théorie était limitée à des décompositions dans un plan (rayonnement dans un demi-espace). L'objet de cette thèse est d'utiliser cette théorie pour décomposer des champs rayonnés ou diffractés dans toutes les directions de l'espace. Ce travail de thèse commence par une étude approfondie de l'influence des paramètres utilisés pour le calcul des coefficients de frame. La mise en oeuvre numérique permet de tester l'efficacité de techniques de troncation et de compression en termes de compromis précision/temps de calcul. Le coeur de la thèse consiste en une méthode originale de partitionnement spectral, utilisant des fonctions de partition de l'unité, qui permet d'utiliser le lancer de faisceaux gaussiens à partir de frames définis dans six plans, pour un rayonnement dans tout l'espace tridimensionnel. La formulation de la méthode est présentée. Elle est appliquée à la décomposition en faisceaux gaussiens du champ rayonné par des antennes théoriques omnidirectionnelles (réseau de dipôles et dipôle demi-onde). Une antenne réaliste sert enfin de cas test pour la mise en œuvre de la décomposition à partir de données expérimentales discrètes

Le présent mémoire comprend un survol des principales méthodes de rendu en demi-tons, de l’analog screening à la recherche binaire directe en passant par l’ordered dither, avec une attention particulière pour la diffusion d’erreur. Ces méthodes seront comparées dans la perspective moderne de la sensibilité à la structure. Une nouvelle méthode de rendu en demi-tons par diffusion d’erreur est présentée et soumise à diverses évaluations. La méthode proposée se veut originale, simple, autant à même de préserver le caractère structurel des images que la méthode à l’état de l’art, et plus rapide que cette dernière par deux à trois ordres de magnitude. D’abord, l’image est décomposée en fréquences locales caractéristiques. Puis, le comportement de base de la méthode proposée est donné. Ensuite, un ensemble minutieusement choisi de paramètres permet de modifier ce comportement de façon à épouser les différents caractères fréquentiels locaux. Finalement, une calibration détermine les bons paramètres à associer à chaque fréquence possible. Une fois l’algorithme assemblé, toute image peut être traitée très rapidement : chaque pixel est attaché à une fréquence propre, cette fréquence sert d’indice pour la table de calibration, les paramètres de diffusion appropriés sont récupérés, et la couleur de sortie déterminée pour le pixel contribue en espérance à souligner la structure dont il fait partie. / This work covers some important methods in the domain of halftoning: analog screening, ordered dither, direct binary search, and most particularly error diffusion. The methods will be compared in the modern perspective of sensitivity to structure. A novel halftoning method is also presented and subjected to various evaluations. It produces images of visual quality comparable to that of the state-of-the-art Structure-aware Halftoning method; at the same time, it is two to three orders of magnitude faster. First is described how an image can be decomposed into its local frequency content. Then, the basic behavior of the proposed method is given. Next, a carefully chosen set of parameters is presented that allow modifications to this behavior, so as to maximize the eventual reactivity to frequency content. Finally, a calibration step determines what values the parameters should take for any local frequency information encountered. Once the algorithm is assembled, any image can be treated very efficiently: each pixel is attached to its dominant frequency, the frequency serves as lookup index to the calibration table, proper diffusion parameters are retrieved, and the determined output color contributes in expectation to underline the structure from which the pixel comes.

<p>We investigate the use of noise degraded facial images in the application of facial expression recognition. In particular, we trained Gabor+SVMclassifiers to recognize facial expressions images with various types of noise. We applied Gaussian noise, Poisson noise, varying levels of salt and pepper noise, and speckle noise to noiseless facial images. Classifiers were trained with images without noise and then tested on the images with noise. Next, the classifiers were trained using images with noise, and then on tested both images that had noise, and images that were noiseless. Finally, classifiers were tested on images while increasing the levels of salt and pepper in the test set. Our results reflected distinct degradation of recognition accuracy. We also discovered that certain types of noise, particularly Gaussian and Poisson noise, boost recognition rates to levels greater than would be achieved by normal, noiseless images. We attribute this effect to the Gaussian envelope component of Gabor filters being sympathetic to Gaussian-like noise, which is similar in variance to that of the Gabor filters. Finally, using linear regression, we mapped a mathematical model to this degradation and used it to suggest how recognition rates would degrade further should more noise be added to the images.</p>

Ce travail de thèse vise à mettre en place un système d'extraction de réseau routier en milieu urbain à partir d'image satellite à très haute résolution. Dans ce contexte, nous avons proposé deux méthodes de localisation de routes. La première approche est fondée sur la procédure de conversion de l'image vers un format vectoriel. L'originalité de cette approche réside dans l'utilisation d'une méthode géométrique pour assurer le passage vers une représentation vectorielle de l'image d'origine et la mise en place d'un formalisme logique fondé sur un ensemble de critères perceptifs permettant le filtrage de l'information inutile et l'extraction des structures linéaires. Dans la deuxième approche, nous avons proposé un algorithme fondé sur la théorie des ondelettes, il met particulièrement en évidence les deux volets multi-résolution et multi-direction. Nous proposons donc une approche de localisation des routes mettant en jeux l'information fréquentielle multi directionnelle issue de la transformée en ondelette Log-Gabor. Dans l'étape de localisation, nous avons présenté deux détecteurs de routes qui exploitent l'information radiométrique, géométrique et fréquentielle. Cependant, ces informations ne permettent pas un résultat exact et précis. Pour remédier à ce problème, un algorithme de suivi s'avère nécessaire. Nous proposons la reconstruction de réseaux routiers par des courbes NURBS. Cette approche est basée sur un ensemble de points de repères identifiés dans la phase de localisation. Elle propose un nouveau concept, que nous avons désigné par NURBSC, basé sur les contraintes géométriques des formes à approximer. Nous connectons les segments de route identifiés afin d'obtenir des tracés continus propres aux routes.

A proper time-frequency transform technique suppresses the blurring and smearing effect of the time-varying Doppler shift on the target image. The conventional target imaging method uses the Fourier transform for extracting the Doppler shift from the received radar pulse. Since the Doppler shift is timevarying for rotating targets, the constructed images will be degraded. In this thesis, the Doppler shift information required for the Range-Doppler image of the target is extracted by using high resolution time-frequency transform techniques. The Wigner-Ville Distribution and the Adaptive Gabor Representation with the Coarse-to-Fine and the Matching Pursuit Search Algorithms are examined techniques for the target imaging system. The modified Matching Pursuit Algorithm, the Matching Pursuit with Reduced Dictionary is proposed which decreases the signal processing time required by the Adaptive Gabor Representation. The Hybrid Matching Pursuit Search Algorithm is also introduced in this thesis work and the Coarse-to-Fine Algorithm and the Matching Pursuit Algorithm are combined for obtaining better representation quality of a signal in the time-frequency domain. The stated techniques are applied on to the sample signals and compared with each other. The application of these techniques in the target imaging system is also performed for the simulated aircrafts.