O reconhecimento facial automático é uma tarefa muito importante para a sociedade moderna, pois possibilita o desenvolvimento de diversas aplicações, tais como o controle de imigração em aeroportos, a autenticação de documentos, etc. Muitas destas aplicações ocorrem em ambientes não controlados, onde as fotos são obtidas com grandes variações de poses e expressões faciais, de iluminação, no uso de maquiagem e acessórios, etc. A tarefa de reconhecimento facial automático em ambientes não controlados é ainda muito desafiadora e tem sido alvo de muitas pesquisas no mundo todo nos últimos anos. Dentro deste contexto, esta tese propõe e implementa um conjunto de novos métodos que visam contribuir para o avanço do estado da arte relacionado a este tema de pesquisa. Inicialmente foi proposto o HBWLF, um filtro para enfatizar as componentes de alta frequência da imagem, sem eliminar as de baixa, realçando assim os diversos detalhes das imagens faciais. Em seguida foi proposta uma versão mais geral deste filtro, o MHBWLF, que considera uma vizinhança circular, ao invés de uma grade regular de 3x3 pixels. O MHBWLF foi aplicado em conjunto com um filtro MOSSE no desenvolvimento de um método para a localização precisa dos centros dos olhos. Aproveitando as características do MHBWLF e inspirado em outros descritores foi proposto um novo descritor, o HBWD. Por fim, foi introduzido um novo método de reconhecimento facial, baseado no HBWD. O método proposto emprega o HBWD para descrever densamente a região periocular e, a fim de reduzir a dimensão dos dados, foi proposto um algoritmo de mapeamento baseado no método de agrupamento k-Means++. Os métodos propostos foram todos avaliados utilizando-se imagens das bases LFW, FGLFW e BioID e os resultados experimentais obtidos mostram que os métodos propostos tem desempenho superior a vários outros métodos estado da arte. / The task of automatic face recognition is very important for modern society and very useful for many kind of applications, such as automatic recognition of credit card users, document authentication, security in big events and others. Further, this is a challenging task when performed in uncontrolled scenarios, which involve great variations in imaging conditions such as illumination, poses and facial expressions, partial occlusion due to hair or glasses, makeup, etc. This thesis first introduces the new High-Boost Weber Local Filter (HBWLF) that emphasizes high-frequency components, without eliminating the low-frequency ones, and thus enhances the details of a face. It is also introduced the new MHBWLF (Multiscale High-Boost Weber Local Filter), which is a multiscale version of HBWLF. A new method for precise eye localization is presented, where a MOSSE filter is used for learning the features enhanced by MHBWLF. This thesis also introduces a new local descriptor called HBWD (High-Boost Weber Descriptor) which combines some features of MHBWLF, SIFT and CS-LMP. Finally, a new method of face recognition is presented. The proposed method basically detects the faces, localizes their eyes and performs a face alignment. After that the region of interest (ROI) is more accurately cropped and described using the new HBWD in a dense sampling scheme (sampling each pixel). A new algorithm, based on the known clustering method k-Means++, reduces the dimensionality of the HBWD descriptors densely applied on each ROI, and produces a signature for the image pair being compared. Finally, a SVM is used to classify the images as a matched or mismatched pair. The proposed methods were evaluated using images from the well-known LFW, FGLFW and BioID databases and the experimental results show that the proposed methods outperform other state-of-the-art approaches.

O reconhecimento facial automático é uma tarefa muito importante para a sociedade moderna, pois possibilita o desenvolvimento de diversas aplicações, tais como o controle de imigração em aeroportos, a autenticação de documentos, etc. Muitas destas aplicações ocorrem em ambientes não controlados, onde as fotos são obtidas com grandes variações de poses e expressões faciais, de iluminação, no uso de maquiagem e acessórios, etc. A tarefa de reconhecimento facial automático em ambientes não controlados é ainda muito desafiadora e tem sido alvo de muitas pesquisas no mundo todo nos últimos anos. Dentro deste contexto, esta tese propõe e implementa um conjunto de novos métodos que visam contribuir para o avanço do estado da arte relacionado a este tema de pesquisa. Inicialmente foi proposto o HBWLF, um filtro para enfatizar as componentes de alta frequência da imagem, sem eliminar as de baixa, realçando assim os diversos detalhes das imagens faciais. Em seguida foi proposta uma versão mais geral deste filtro, o MHBWLF, que considera uma vizinhança circular, ao invés de uma grade regular de 3x3 pixels. O MHBWLF foi aplicado em conjunto com um filtro MOSSE no desenvolvimento de um método para a localização precisa dos centros dos olhos. Aproveitando as características do MHBWLF e inspirado em outros descritores foi proposto um novo descritor, o HBWD. Por fim, foi introduzido um novo método de reconhecimento facial, baseado no HBWD. O método proposto emprega o HBWD para descrever densamente a região periocular e, a fim de reduzir a dimensão dos dados, foi proposto um algoritmo de mapeamento baseado no método de agrupamento k-Means++. Os métodos propostos foram todos avaliados utilizando-se imagens das bases LFW, FGLFW e BioID e os resultados experimentais obtidos mostram que os métodos propostos tem desempenho superior a vários outros métodos estado da arte. / The task of automatic face recognition is very important for modern society and very useful for many kind of applications, such as automatic recognition of credit card users, document authentication, security in big events and others. Further, this is a challenging task when performed in uncontrolled scenarios, which involve great variations in imaging conditions such as illumination, poses and facial expressions, partial occlusion due to hair or glasses, makeup, etc. This thesis first introduces the new High-Boost Weber Local Filter (HBWLF) that emphasizes high-frequency components, without eliminating the low-frequency ones, and thus enhances the details of a face. It is also introduced the new MHBWLF (Multiscale High-Boost Weber Local Filter), which is a multiscale version of HBWLF. A new method for precise eye localization is presented, where a MOSSE filter is used for learning the features enhanced by MHBWLF. This thesis also introduces a new local descriptor called HBWD (High-Boost Weber Descriptor) which combines some features of MHBWLF, SIFT and CS-LMP. Finally, a new method of face recognition is presented. The proposed method basically detects the faces, localizes their eyes and performs a face alignment. After that the region of interest (ROI) is more accurately cropped and described using the new HBWD in a dense sampling scheme (sampling each pixel). A new algorithm, based on the known clustering method k-Means++, reduces the dimensionality of the HBWD descriptors densely applied on each ROI, and produces a signature for the image pair being compared. Finally, a SVM is used to classify the images as a matched or mismatched pair. The proposed methods were evaluated using images from the well-known LFW, FGLFW and BioID databases and the experimental results show that the proposed methods outperform other state-of-the-art approaches.

L'oculométrie est un ensemble de techniques dédié à enregistrer et analyser les mouvements oculaires. Dans cette thèse, je présente l'étude, la conception et la mise en œuvre d'un système oculométrique numérique, non-intrusif permettant d'analyser les mouvements oculaires en temps réel avec une webcam à distance et sans lumière infra-rouge. Dans le cadre de la réalisation, le système oculométrique proposé se compose de quatre modules: l'extraction des caractéristiques, la détection et le suivi des yeux, l'analyse de la variété des mouvements des yeux à partir des images et l'estimation du regard par l'apprentissage. Nos contributions reposent sur le développement des méthodes autour de ces quatre modules: la première réalise une méthode hybride pour détecter et suivre les yeux en temps réel à partir des techniques du filtre particulaire, du modèle à formes actives et des cartes des yeux (EyeMap); la seconde réalise l'extraction des caractéristiques à partir de l'image des yeux en utilisant les techniques des motifs binaires locaux; la troisième méthode classifie les mouvements oculaires selon la variété générée par le Laplacian Eigenmaps et forme un ensemble de données d'apprentissage; enfin, la quatrième méthode calcul la position du regard à partir de cet ensemble d'apprentissage. Nous proposons également deux méthodes d'estimation:une méthode de la régression par le processus gaussien et un apprentissage semi-supervisé et une méthode de la catégorisation par la classification spectrale (spectral clustering). Il en résulte un système complet, générique et économique pour les applications diverses dans le domaine de l'oculométrie. / Gaze tracker offers a powerful tool for diverse study fields, in particular eye movement analysis. In this thesis, we present a new appearance-based real-time gaze tracking system with only a remote webcam and without infra-red illumination. Our proposed gaze tracking model has four components: eye localization, eye feature extraction, eye manifold learning and gaze estimation. Our research focuses on the development of methods on each component of the system. Firstly, we propose a hybrid method to localize in real time the eye region in the frames captured by the webcam. The eye can be detected by Active Shape Model and EyeMap in the first frame where eye occurs. Then the eye can be tracked through a stochastic method, particle filter. Secondly, we employ the Center-Symmetric Local Binary Patterns for the detected eye region, which has been divided into blocs, in order to get the eye features. Thirdly, we introduce manifold learning technique, such as Laplacian Eigen-maps, to learn different eye movements by a set of eye images collected. This unsupervised learning helps to construct an automatic and correct calibration phase. In the end, as for the gaze estimation, we propose two models: a semi-supervised Gaussian Process Regression prediction model to estimate the coordinates of eye direction; and a prediction model by spectral clustering to classify different eye movements. Our system with 5-points calibration can not only reduce the run-time cost, but also estimate the gaze accurately. Our experimental results show that our gaze tracking model has less constraints from the hardware settings and it can be applied efficiently in different real-time applications.

Ophthalmology is a branch of medicine that deals with the anatomy, physiology and diseases of the eyeball and orbit. An ophthalmic device for the acquirement and recognition of a human eye characteristics was created by researchers from Faculty of Information Technology. This device can be used either for biometric purposes or for medical purposes as a support diagnostic device. To achieve proper functionality, device must be able to adjust platform position in order to align optic camera with patient's eye.  The main focus of this thesis is to design and implement an algorithm for eye centre localisation based on images of the patient's face. The first part of this thesis describes general methods for eye localisation and proposed solution. To achieve requested accuracy, combination of multiple methods is used with adjusted parameters based on platform's features. The second part describes implementation of proposed solution as well as platform control. Multiple databases were used for training and testing of the algorithm. The third part summarises performed experiments. The proposed algorithm was implemented in the C++ language, using OpenCV library. Accuracy and speed of proposed algorithm are suitable for developed platform. In the end, the results are discussed and further improvements are proposed.

Ophthalmology is the branch of medicine that deals with the anatomy, physiology and diseases of the eyeball and orbit. An ophthalmic device for acquirement and recognition of human eye characteristics was created by researchers from Faculty of Information Technology. This device can be used either for biometric purposes or for medical purposes as a support diagnostic device. To achieve proper functionality, device must be able to adjust platform position in order to align optic camera with patient's eye.  The main focus of this work is to design and implement an algorithm for eye centre localisation based on images of the patient's face. The first part of this thesis describes general methods for eye localisation and proposed solution. To achieve desired accuracy, combination of multiple methods is used with adjusted parameters based on platform's features. The second part describes implementation of proposed solution as well as platform control. Multiple databases were used for training and testing of the algorithm and the third part summarises performed experiments. The proposed algorithm was implemented in the C++ language, using OpenCV library. Accuracy and speed of proposed algorithm are suitable for developed platform. In the end, the results are discussed and further improvements are proposed.