Dans cette thèse, nous nous intéressons à la classification des images de textures avec aucune connaissance a priori sur les conditions de numérisation. Cette classification selon des définitions pré-établies de matériaux repose sur des algorithmes qui extraient des descripteurs visuels.A cette fin, nous introduisons tout d'abord une variante de descripteurs par motifs binaires locaux (Local Binary Patterns).Dans cette proposition, une approche statistique est suivie pour représenter les textures statiques.Elle incorpore la quantité d'information complémentaire des niveaux de gris des images dans des opérateurs basés LBP.Nous avons nommé cette nouvelle méthode "Completed Local Entropy Binary Patterns (CLEBP)".CLEBP capture la distribution des relations entre les mesures statistiques des données aléatoires d'une image, l'ensemble étant calculé pour tous les pixels au sein d'une structure locale.Sans la moindre étape préalable d'apprentissage, ni de calibration automatique, les descriptions CLEBP contiennent à la fois des informations locales et globales des textures, tout en étant robustes aux variations externes.En outre, nous utilisons le filtrage inspiré par la biologie, ou biologically-inspired filtering (BF), qui simule la rétine humaine via une phase de prétraitement.Nous montrons que notre approche est complémentaire avec les LBP conventionnels, et les deux combinés offrent de meilleurs résultats que l'une des deux méthodes seule.Les résultats expérimentaux sur quatre bases de texture, Outex, KTH-TIPS-2b, CURet, et UIUC montrent que notre approche est plus performante que les méthodes actuelles.Nous introduisons également un cadre formel basé sur une combinaison de descripteurs pour la classification de textures.Au sein de ce cadre, nous combinons des descripteurs LBP invariants en rotation et en échelle, et de faible dimension, avec les réseaux de dispersion, ou scattering networks (ScatNet).Les résultats expérimentaux montrent que l'approche proposée est capable d'extraire des descripteurs riches à de nombreuses orientations et échelles.Les textures sont modélisées par une concaténation des codes LBP et valeurs moyennes des coefficients ScatNet.Nous proposons également d'utiliser le filtrage inspiré par la biologie, ou biologically-inspired filtering (BF), pour améliorer la resistance des descripteurs LBP.Nous démontrons par l'expérience que ces nouveaux descripteurs présentent de meilleurs résultats que les approches usuelles de l'état de l'art.Ces résultats sont obtenus sur des bases réelles qui contiennent de nombreuses avec des variations significatives.Nous proposons aussi un nouveau réseau conçu par l'expertise appelé réseaux de convolution normalisée, ou normalized convolution network.Celui-ci est inspiré du modèle des ScatNet, auquel deux modifications ont été apportées.La première repose sur l'utilisation de la convolution normalisé en lieu et place de la convolution standard.La deuxième propose de remplacer le calcul de la valeur moyenne des coefficients du réseaux par une agrégation avec la méthode des vecteurs de Fisher.Les expériences montrent des résultats compétitifs sur de nombreuses bases de textures.Enfin, tout au long de cette thèse, nous avons montré par l'expérience qu'il est possible d'obtenir de très bons résultats de classification en utilisant des techniques peu coûteuses en ressources. / This thesis investigates the complementary approaches for classifying texture images.The thesis begins by proposing a Local Binary Pattern (LBP) variant for efficient texture classification.In this proposed method, a statistical approach to static texture representation is developed. It incorporates the complementary quantity information of image intensity into the LBP-based operators. We name our LBP variant `the completed local entropy binary patterns (CLEBP)'. CLEBP captures the distribution of the relationships between statistical measures of image data randomness, calculated over all pixels within a local structure. Without any pre-learning process and any additional parameters to be learned, the CLEBP descriptors convey both global and local information about texture while being robust to external variations. Furthermore, we use biologically-inspired filtering (BF) which simulates the performance of human retina as preprocessing technique. It is shown that our approach and the conventional LBP have the complementary strength and that by combining these algorithms, one obtains better results than either of them considered separately. Experimental results on four large texture databases show that our approach is more efficient than contemporary ones.We then introduce a framework which is a feature combination approach to the problem of texture classification. In this framework, we combine Local Binary Pattern (LBP) features with low dimensional, rotation and scale invariant counterparts, the handcrafted scattering network (ScatNet). The experimental results show that the proposed approach is capable of extracting rich features at multiple orientations and scales. Textures are modeled by concatenating histogram of LBP codes and the mean values of ScatNet coefficients. Then, we propose using Biological Inspired Filtering (BF) preprocessing technique to enhance the robustness of LBP features. We have demonstrated by experiment that the novel features extracted from the proposed framework achieve superior performance as compared to their traditional counterparts when benchmarked on real-world databases containing many classes with significant imaging variations.In addition, we propose a novel handcrafted network called normalized convolution network. It is inspired by the model of ScatNet with two important modification. Firstly, normalized convolution substitute for standard convolution in ScatNet model to extract richer texture features. Secondly, Instead of using mean values of the network coefficients, Fisher vector is exploited as an aggregation method. Experiments show that our proposed network gains competitive classification results on many difficult texture benchmarks.Finally, throughout the thesis, we have proved by experiments that the proposed approaches gain good classification results with low resource required.

A geometallurgical model is the combination of a spatial model representing an ore deposit and a process model representing the comminution and concentration steps in beneficiation. The process model itself usually consists of several unit models. Each of these unit models operates at a given level of detail in material characterization - from bulk chemical elements, elements by size, bulk minerals and minerals by size to the liberation level that introduces particles as the basic entity for simulation (Paper 1). In current state-of-the-art process simulation, few unit models are defined at the particle level because these models are complex to design at a more fundamental level of detail, liberation data is hard to measure accurately and large computational power is required to process the many particles in a flow sheet. Computational cost is a consequence of the intrinsic complexity of the unit models. Mineral liberation data depends on the quality of the sampling and the polishing, the settings and stability of the instrument and the processing of the data. This study introduces new tools to simulate a population of mineral particles based on intrinsic characteristics of the feed ore. Features are extracted at the meso-textural level (drill cores) (Paper 2), put in relation to their micro-textures before breakage and after breakage (Paper 3). The result is a population of mineral particles stored in a file format compatible to import into process simulation software. The results show that the approach is relevant and can be generalized towards new characterization methods. The theory of image representation, analysis and ore texture simulation is briefly introduced and linked to 1-point, 2-point, and multiple-point methods from spatial statistics. A breakage mechanism is presented as a cellular automaton. Experimental data and examples are taken from a copper-gold deposit with a chalcopyrite flotation circuit, an iron ore deposit with a magnetic separation process. This study is covering a part of a larger research program, PREP (Primary resource efficiency by enhanced prediction). / PREP

A classificação de texturas coloridas sob diferentes condições de iluminação é um desafio na área de visão computacional, e depende da eficiência dos descritores de textura em capturar características que sejam discriminantes independentemente das propriedades da fonte de luz incidente sobre o objeto. Visando melhorar o processo de classificação de texturas coloridas iluminadas com diferentes fontes de luz, este trabalho propõe três novos descritores, nomeados Opponent Color Local Mapped Pattern (OCLMP), que combina o descritor de texturas por padrões locais mapeados (Local Mapped Pattern - LMP) com a teoria de cores oponentes; Color Intensity Local Mapped Pattern (CILMP), que extrai as informações de cor e textura de maneira integrada, levando em consideração a textura da cor, combinando estas informações com características da luminância da textura em uma análise multiresolução; e Extended Color Local Mapped Pattern (ECLMP), que utiliza dois operadores para extrair informações de cor e textura de forma integrada (textura da cor) combinadas com informações apenas de textura (sem cor) de uma imagem. Todos esses novos descritores propostos são paramétricos e, sendo o ajuste ótimo de seus parâmetros não trivial, o processo exige um tempo excessivo de computação. Portanto, foi proposto nesta tese a utilização de algoritmos genéticos para o ajuste automático dos parâmetros. A avaliação dos descritores propostos foi realizada em duas bases de dados de texturas coloridas com variação de iluminação: RawFooT (Raw Food Texture Database) e KTH-TIPS- 2b (Textures under varying Illumination, Pose and Scale Database), utilizando-se um classificador. Os resultados experimentais mostraram que os descritores propostos são mais robustos à variação de iluminação do que outros decritores de textura comumente utilizados na literatura. Os descritores propostos apresentaram um desempenho superior aos descritores comparados em 15% na base de dados RawFooT e 4% na base de dados KTH-TIPS-2b. / Color texture classification under varying illumination remains a challenge in the computer vision field, and it greatly relies on the efficiency at which the texture descriptors capture discriminant features, independent of the illumination condition. The aim of this thesis is to improve the classification of color texture acquired with varying illumination sources. We propose three new color texture descriptors, namely: the Opponent Color Local Mapped Pattern (OCLMP), which combines a local methodology (LMP) with the opponent colors theory, the Color Intensity Local Mapped Pattern (CILMP), which extracts color and texture information jointly, in a multi-resolution fashion, and the Extended Color Local Mapped Pattern (ECLMP), which applies two operators to extract color and texture information jointly as well. As the proposed methods are based on the LMP algorithm, they are parametric functions. Finding the optimal set of parameters for the descriptor can be a cumbersome task. Therefore, this work proposes the use of genetic algorithms to automatically adjust the parameters. The methods were assessed using two data sets of textures acquired using varying illumination sources: the RawFooT (Raw Food Texture Database), and the KTH-TIPS-2b (Textures under varying Illumination, Pose and Scale Database). The experimental results show that the proposed descriptors are more robust to variations to the illumination source than other methods found in the literature. The improvement on the accuracy was higher than 15% on the RawFoot data set, and higher than 4% on the KTH-TIPS-2b data set.

Classificação de imagens de textura, especialmente aquelas com mudanças significativas de rotação, iluminação, escala e ponto de vista, é um problema fundamental e desafiador na área de visão computacional. Esta tese propõe dois descritores de imagem simples, porém eficientes, chamados de Sampled Local Mapped Pattern Magnitude (SLMP_M) e Completed Local Mapped Pattern (CLMP) aplicados na classificação de textura. Os descritores propostos são parte de um aprimoramento do descritor Local Mapped Pattern (LMP) para trabalhar de maneira eficiente com imagens de textura rotacionadas. Os métodos propostos necessitam de um pré-ajuste de parâmetros que utiliza o método de otimização por enxame de partículas, e são discriminativos e robustos para a descrição de texturas rotacionadas em ângulos arbitrários. Para a validação dos descritores propostos duas bases de imagens são utilizadas, Kylberg Sintorn Rotation Dataset e Brodatz Texture Rotation Dataset, uma nova base de dados desenvolvida pela autora, formada por imagens de texturas rotacionadas do Álbum de Brodatz. As duas bases contêm imagens de texturas naturais que foram rotacionadas fisicamente no momento da captura e rotacionadas por processos computacionais. É feita também uma avaliação da influência de métodos de interpolação no processo de rotação das imagens e são comparados com diferentes descritores presentes na literatura. Cinco métodos de interpolação são investigados: Lanczos, B-spline, Cúbica, Linear e Nearest Neighbor. Os resultados experimentais demonstram que os descritores propostos nesta tese superam o desempenho dos descritores Completed Local Binary Pattern (CLBP), e dos descritores que combinam a versão generalizada das características de Fourier com variações do descritor Local Binary Pattern (LBP), LBPDFT, ILBPDFT, LTPDFT e ILTPDFT. Os resultados também demonstram que a escolha do método de interpolação no processo de rotação das imagens influencia na capacidade de reconhecimento. / Texture image classification, especially those with significant changes of rotation, illumination, scale and point of view, is a fundamental and challenging problem in the field of computer vision. This thesis proposes two simple, but efficient, image descriptors called Sampled Local Mapped Pattern Magnitude (SLMP_M) and Completed Local Mapped Pattern (CLMP) applied in texture classification. The proposed descriptors are part of an enhancement to the Local Mapped Pattern (LMP) descriptor to work efficiently with rotated texture images. The descriptors proposed requires a parameter preset by the particle swarm optimization method, they are discriminating and robust for the description of rotated textures at arbitrary angles. For the validation of the proposed descriptors two image datasets are used: Kylberg Sintorn Rotation Dataset and Brodatz Texture Rotation Dataset, a new texture dataset introduced, which contains rotated texture images from Brodatzs Album. Both databases contain images of natural textures that have been rotated by Hardware and computational procedures. An evaluation of the influence of interpolation methods on the image rotation process is also presented and compared with different descriptors in the literature. Five interpolation methods are investigated: Lanczos, B-spline, Cubic, Linear and Nearest Neighbor. The experimental results show that the descriptors proposed in this thesis outperform the performance of the Completed Local Binary Pattern (CLBP) descriptors, and the descriptors that combine the generalized version of the Fourier characteristics with variations of the descriptor Local Binary Pattern (LBP), LBPDFT, ILBDFT, LTPDFT e ILTPDFT compared. The results also prove that the selection of the interpolation method in the image rotation process influences the recognition capability.

A classificação de texturas coloridas sob diferentes condições de iluminação é um desafio na área de visão computacional, e depende da eficiência dos descritores de textura em capturar características que sejam discriminantes independentemente das propriedades da fonte de luz incidente sobre o objeto. Visando melhorar o processo de classificação de texturas coloridas iluminadas com diferentes fontes de luz, este trabalho propõe três novos descritores, nomeados Opponent Color Local Mapped Pattern (OCLMP), que combina o descritor de texturas por padrões locais mapeados (Local Mapped Pattern - LMP) com a teoria de cores oponentes; Color Intensity Local Mapped Pattern (CILMP), que extrai as informações de cor e textura de maneira integrada, levando em consideração a textura da cor, combinando estas informações com características da luminância da textura em uma análise multiresolução; e Extended Color Local Mapped Pattern (ECLMP), que utiliza dois operadores para extrair informações de cor e textura de forma integrada (textura da cor) combinadas com informações apenas de textura (sem cor) de uma imagem. Todos esses novos descritores propostos são paramétricos e, sendo o ajuste ótimo de seus parâmetros não trivial, o processo exige um tempo excessivo de computação. Portanto, foi proposto nesta tese a utilização de algoritmos genéticos para o ajuste automático dos parâmetros. A avaliação dos descritores propostos foi realizada em duas bases de dados de texturas coloridas com variação de iluminação: RawFooT (Raw Food Texture Database) e KTH-TIPS- 2b (Textures under varying Illumination, Pose and Scale Database), utilizando-se um classificador. Os resultados experimentais mostraram que os descritores propostos são mais robustos à variação de iluminação do que outros decritores de textura comumente utilizados na literatura. Os descritores propostos apresentaram um desempenho superior aos descritores comparados em 15% na base de dados RawFooT e 4% na base de dados KTH-TIPS-2b. / Color texture classification under varying illumination remains a challenge in the computer vision field, and it greatly relies on the efficiency at which the texture descriptors capture discriminant features, independent of the illumination condition. The aim of this thesis is to improve the classification of color texture acquired with varying illumination sources. We propose three new color texture descriptors, namely: the Opponent Color Local Mapped Pattern (OCLMP), which combines a local methodology (LMP) with the opponent colors theory, the Color Intensity Local Mapped Pattern (CILMP), which extracts color and texture information jointly, in a multi-resolution fashion, and the Extended Color Local Mapped Pattern (ECLMP), which applies two operators to extract color and texture information jointly as well. As the proposed methods are based on the LMP algorithm, they are parametric functions. Finding the optimal set of parameters for the descriptor can be a cumbersome task. Therefore, this work proposes the use of genetic algorithms to automatically adjust the parameters. The methods were assessed using two data sets of textures acquired using varying illumination sources: the RawFooT (Raw Food Texture Database), and the KTH-TIPS-2b (Textures under varying Illumination, Pose and Scale Database). The experimental results show that the proposed descriptors are more robust to variations to the illumination source than other methods found in the literature. The improvement on the accuracy was higher than 15% on the RawFoot data set, and higher than 4% on the KTH-TIPS-2b data set.

Texture is one of the most important characteristics used in computer vision and image processing applications. In this thesis, a new texture classification and retrieval method is proposed for texture analysis applications. The technique makes use of the random neural network model and it is supervised. The main aim is to represent textures with parameters which are the random neural network weights and classify and retrieve textures using this texture definition. The network has neurons that correspond to each image pixel, and the neurons are connected according to neighboring relationship between pixels. The method is tested on artificial images produced by using Brodatz album and texture blocks cut from remotely sensed imagery.

Texture analysis is an important problem in machine vision, with applications in many fields including medical imaging, remote sensing (SAR), automated flaw detection in various products, and document analysis to name but a few. Over the last four decades many techniques for the analysis of textured images have been proposed in the literature for the purposes of classification, segmentation, synthesis and compression. Such approaches include analysis the properties of individual texture elements, using statistical features obtained from the grey-level values of the image itself, random field models, and multichannel filtering. The wavelet transform, a unified framework for the multiresolution decomposition of signals, falls into this final category, and allows a texture to be examined in a number of resolutions whilst maintaining spatial resolution. This thesis explores the use of the wavelet transform to the specific task of texture classification and proposes a number of improvements to existing techniques, both in the area of feature extraction and classifier design. By applying a nonlinear transform to the wavelet coefficients, a better characterisation can be obtained for many natural textures, leading to increased classification performance when using first and second order statistics of these coefficients as features. In the area of classifier design, a combination of an optimal discriminate function and a non-parametric Gaussian mixture model classifier is shown to experimentally outperform other classifier configurations. By modelling the relationships between neighbouring bands of the wavelet trans- form, more information regarding a texture can be obtained. Using such a representation, an efficient algorithm for the searching and retrieval of textured images from a database is proposed, as well as a novel set of features for texture classification. These features are experimentally shown to outperform features proposed in the literature, as well as provide increased robustness to small changes in scale. Determining the script and language of a printed document is an important task in the field of document processing. In the final part of this thesis, the use of texture analysis techniques to accomplish these tasks is investigated. Using maximum a posterior (MAP) adaptation, prior information regarding the nature of script images can be used to increase the accuracy of these methods. Novel techniques for estimating the skew of such documents, normalising text block prior to extraction of texture features and accurately classifying multiple fonts are also presented.

Classificação de imagens de textura, especialmente aquelas com mudanças significativas de rotação, iluminação, escala e ponto de vista, é um problema fundamental e desafiador na área de visão computacional. Esta tese propõe dois descritores de imagem simples, porém eficientes, chamados de Sampled Local Mapped Pattern Magnitude (SLMP_M) e Completed Local Mapped Pattern (CLMP) aplicados na classificação de textura. Os descritores propostos são parte de um aprimoramento do descritor Local Mapped Pattern (LMP) para trabalhar de maneira eficiente com imagens de textura rotacionadas. Os métodos propostos necessitam de um pré-ajuste de parâmetros que utiliza o método de otimização por enxame de partículas, e são discriminativos e robustos para a descrição de texturas rotacionadas em ângulos arbitrários. Para a validação dos descritores propostos duas bases de imagens são utilizadas, Kylberg Sintorn Rotation Dataset e Brodatz Texture Rotation Dataset, uma nova base de dados desenvolvida pela autora, formada por imagens de texturas rotacionadas do Álbum de Brodatz. As duas bases contêm imagens de texturas naturais que foram rotacionadas fisicamente no momento da captura e rotacionadas por processos computacionais. É feita também uma avaliação da influência de métodos de interpolação no processo de rotação das imagens e são comparados com diferentes descritores presentes na literatura. Cinco métodos de interpolação são investigados: Lanczos, B-spline, Cúbica, Linear e Nearest Neighbor. Os resultados experimentais demonstram que os descritores propostos nesta tese superam o desempenho dos descritores Completed Local Binary Pattern (CLBP), e dos descritores que combinam a versão generalizada das características de Fourier com variações do descritor Local Binary Pattern (LBP), LBPDFT, ILBPDFT, LTPDFT e ILTPDFT. Os resultados também demonstram que a escolha do método de interpolação no processo de rotação das imagens influencia na capacidade de reconhecimento. / Texture image classification, especially those with significant changes of rotation, illumination, scale and point of view, is a fundamental and challenging problem in the field of computer vision. This thesis proposes two simple, but efficient, image descriptors called Sampled Local Mapped Pattern Magnitude (SLMP_M) and Completed Local Mapped Pattern (CLMP) applied in texture classification. The proposed descriptors are part of an enhancement to the Local Mapped Pattern (LMP) descriptor to work efficiently with rotated texture images. The descriptors proposed requires a parameter preset by the particle swarm optimization method, they are discriminating and robust for the description of rotated textures at arbitrary angles. For the validation of the proposed descriptors two image datasets are used: Kylberg Sintorn Rotation Dataset and Brodatz Texture Rotation Dataset, a new texture dataset introduced, which contains rotated texture images from Brodatzs Album. Both databases contain images of natural textures that have been rotated by Hardware and computational procedures. An evaluation of the influence of interpolation methods on the image rotation process is also presented and compared with different descriptors in the literature. Five interpolation methods are investigated: Lanczos, B-spline, Cubic, Linear and Nearest Neighbor. The experimental results show that the descriptors proposed in this thesis outperform the performance of the Completed Local Binary Pattern (CLBP) descriptors, and the descriptors that combine the generalized version of the Fourier characteristics with variations of the descriptor Local Binary Pattern (LBP), LBPDFT, ILBDFT, LTPDFT e ILTPDFT compared. The results also prove that the selection of the interpolation method in the image rotation process influences the recognition capability.

Textured motion - generally known as dynamic or temporal texture - is a popular research area for synthesis, segmentation and recognition. Dynamic texture is a spatially repetitive, time-varying visual pattern that forms an image sequence with certain temporal stationarity. In dynamic texture, the notion of self-similarity central to conventional image texture is extended to the spatiotemporal domain. Dynamic textures are typically videos of processes, such as waves, smoke, fire, a flag blowing in the wind, a moving escalator, or a walking crowd. Creation of synthetic frames is a key issue especially for movie screen industry to enrich their scenes from a white screen into a shining reality. In robotics world, for example an autonomous vehicle must decide what is traversable terrain (e.g. grass) and what is not (e.g. water). This problem can be addressed by classifying portions of the image into a number of categories, for instance grass, dirt, bushes or water. If these parts are identifiable, then segmentation and recognition of these textures results with an efficient path planning for the autonomous vehicle. In this thesis, we aimed to characterize these textured motions like mentioned above. We tried to implement several known techniques and compared the results.

A inspeção visual automática é uma tarefa importante para a produtividade industrial. Ela pode ser aplicada em controle de qualidade para substituir operadores humanos em trabalhos perigosos ou repetitivos. O estágio de classificação em controle de qualidade da produção industrial é freqüentemente baseado no conhecimento humano. Portanto, torna-se importante alimentar um sistema visual automático com dados nebulosos ou ambíguos. Um sistema \"neuro-fuzy\" é uma forma adequada de implementar isto. O trabalho contribui na área tecnológica de inspeção visual com o desenvolvimento de uma nova abordagem para avaliação da qualidade de placas de madeira utilizadas na fabricação de lápis. Outra contribuição foi a divisão do vetor de características, fazendo com que cada característica específica seja tratada em uma rede neural própria. O método é baseado em duas redes neurais, cada uma tratando com apenas uma característica de entrada. Os resultados das redes neurais são combinados através de lógica nebulosa (\"fuzzy) fornecendo um sistema com maior poder discriminante do que aqueles que utilizam métodos tradicionais. O sistema se caracteriza por ser ágil, repetitivo, com um padrão de classificação definido e por possuir baixo custo. / Automatic visual inspection is an important task for industrial productivity. It could be applied for quality control or for replacing manual work under dangerous or repetitive activity. The classification stage in quality control of the industrial production is often based on the human knowledge. It seems, therefore, to be a great concern to supply an automated visual inspection system with fuzzy or ambiguous data. The Neuro-Fuzzy system is a good way to do this. The objective of this work is to develop a new approach for the classification of wooden plates used in the pencil production. This new method is based on two neural networks, each one working with just an input feature. The results of neural networks are combined through fuzzy logic giving the system a greater discriminating power than those that use traditional methods. The proposed method is characterized by being agile, repetitive, with a defined classification pattern and having low cost.