A textura de uma imagem apresenta informações importantes sobre as características de um objeto. Usar essa informação para reconhecimento de padrões vem sendo uma tarefa bastante pesquisada na área de processamento de imagens e aplicado em atividades como indústria têxtil, biologia, análise de imagens médicas, imagens de satélite, análise de peças industriais, entre outros. Muitos pesquisadores focam em criar mecanismos que convertam a imagem em um vetor de características a fim de utilizar um classificador sobre esse vetores. No entanto, as imagens podem ser transformadas para que que características peculiares sejam evidenciadas fazendo com que extratores de características já existentes explorem melhor as imagens. Esse trabalho tem como objetivo estudar a influência da aplicação de métodos de pré-processamento em imagens de textura para a posterior análise das imagens. Os métodos escolhidos são seis: difusão isotrópica, difusão anisotrópica clássica, dois métodos de regularização da difusão anisotrópica, um método de difusão morfológica e a transformada de distância. Além disso, os métodos foram aliados a sete descritores já conhecidos da literatura para que as características das imagens tranformadas sejam extraídas. Resultados mostram um aumento significativo no desempenho dos classificadores KNN e Naive Bayes quando utilizados nas imagens transformadas de quatro bases de textura: Brodatz, Outex, Usptex e Vistex. / The texture of an image plays an important source of information of the image content. The use of this information to pattern recognition became very popular in image processing area and has applications such in textile industry, biology, medical image analysis, satelite images analysis, industrial equipaments analysis, among others. Many researchers focus on creating different methods to convert the input image to a feature vector to the able to classify the image based on these vectors. However, images can be modified in different ways such that important features are enhanced. Therefore, descriptors are able to extract features easily to perform a better representation of the image. This project aims to apply six different preprocessing methods to analyze their power of enhancement on the texture extraction. The methods are: isotropic diffusion, the classic anisotropic diffusion, two regularizations of the anisotropic diffusion, a morphologic diffusion and the distance transform. To extract the features of these modified images, seven texture analysis algorithms are used along KNN and Naive Bayes to classify the textures. Results show a significant increase when datasets Brodatz, Vistex, Usptex and Outex are transformed prior to texture analysis and classification.

Submitted by FERNANDO PEREIRA DOS SANTOS null (fernando_persan@hotmail.com) on 2016-06-30T21:35:05Z No. of bitstreams: 1 EstudoComparativo_FernandoPereiraSantos.pdf: 17823712 bytes, checksum: fd7d1643bbacb3313ced8e742fda123f (MD5) / Approved for entry into archive by Ana Paula Grisoto (grisotoana@reitoria.unesp.br) on 2016-07-04T12:57:57Z (GMT) No. of bitstreams: 1 santos_fp_me_sjrp.pdf: 17823712 bytes, checksum: fd7d1643bbacb3313ced8e742fda123f (MD5) / Made available in DSpace on 2016-07-04T12:57:57Z (GMT). No. of bitstreams: 1 santos_fp_me_sjrp.pdf: 17823712 bytes, checksum: fd7d1643bbacb3313ced8e742fda123f (MD5) Previous issue date: 2016-06-16 / Neste trabalho é apresentada uma nova metodologia para a análise e classificação de lesões de pele. O modelo proposto foi dividido em quatro etapas que compreende o pré-processamento das imagens, a segmentação do objeto desejado, a extração de características e a classificação das lesões. Na etapa de pré-processamento, aplica-se o modelo de cor RGB, a quantização de cores e o filtro de difusão anisotrópica. Na segmentação, a imagem suavizada é submetida à operação de fechamento da morfologia matemática, estimativa de peso nos arcos do grafo e à transformada imagem-floresta com apenas duas sementes. A extração de características foi baseada na aplicação da regra ABCD. Na assimetria foi aplicado o conceito de razão de perpendiculares sobre a maior diagonal obtida e, para a borda, o produto vetorial e o ponto de inflexão foram implementados para fornecer o porcentual de curvatura do contorno. Para a cor, valores de média, variância, desvio padrão, homogeneidade e contraste foram calculados. Para a estrutura diferencial foi desenvolvido a dimensão fractal e a energia. Na última etapa, classificação, a floresta de caminhos ótimos foi utilizada. Os resultados da classificação são apresentados por malignidade, quando todos os tipos de lesões estão juntos, e por categorias, quando os tipos de lesão são agrupados dois a dois. Para obter o modelo proposto foram efetuados diversos testes com modelos de cor diferentes, forma de aplicação da quantização, diferenciação no cálculo da quantidade de iterações do filtro de difusão anisotrópica e possibilidade de não aplicar a morfologia matemática. / In this paper a new methodology for the analysis and classification of skin lesions is presented. The suggested model is divided into four steps which comprise the pre processing of images, segmentation of the desired object, feature extraction and lesions classification. In the pre processing step, it is applied the RGB color model, color quantization and anisotropic diffusion filter. In segmentation, the smoothed image is submitted to the closing operation of mathematical morphology, arc-weight estimation in the graph and the image-foresting transform with only two seeds. The feature extraction is based on the application of ABCD rule. In asymmetry was applied the perpendicular ratio concept on the greater diagonal obtained and, to the border, the vector product and the inflection point were implemented to provide the contour curvature percentage. For color, average values, variance, standard deviation, homogeneity and contrast were calculated. For differential structure was developed fractal dimension and energy. In the last stage, classification, optimum-path forest was used. The classification results are presented by malignancy, when all types of lesions are together, and by categories, when the types of lesions are grouped two by two. For the model were performed several tests with different color models, the form of application the quantization, differentiation in the calculation of the quantity of filter iterations of anisotropic diffusion and the possibility of not applying the mathematical morphology.

Diploma thesis is about image sensor security. Goal of the thesis was study data integrity gained from the image sensors. Proposed method is about source camera identification from noise characteristics in image sensors. Research was about influence of denoising algorithms applied to digital images, which was acquired from 15 different image sensors. Finally the statistical evaluation had been done from computed results.

This diploma thesis deals with the methods of despeckling in ultrasound images. Ultrasound imaging and related artifacts are described in more details. Ultrasound imaging has its pros and cons, where speckle noise is a disadvantage to be solved. Models of origin of this specific noise are referred too. Practical part of this thesis aims on filtering speckled images by basic and advanced filtering methods as are linear filtering, median filtering, application of Frost filter, QGDCT, geometric filtering, anisotropic diffusion filtering and filtering based on wavelet transformation. Results are compared on the basis of objective criteria.

The aim of this work was to establish a comprehensive processing pipeline of cervical spinal cord HARDI dMRI data and T2-weighted anatomical MRI images in high-resolution. In the research part we provide description of anatomical data processing, theoretical background of dMRI, description of current approaches to 3D anisotropic diffusion estimation as well as current imaging methods of spinal cord axonal bundles. As one of the first in the world, we are investigating multiple-direction diffusion models for human in-vivo spinal cord white matter minority bundles imaging. We designed our own processing pipeline utilizing Spinal Cord Toolbox (SCT), FSL, in-house developer scripts and TORQUE-based batch system for grid computation, tested on real data from cervical spinal cord area between segments C4-C6 from 26 healthy volunteers. Designed processing pipeline with one non-automatic step, works from pre-processing to parcelation of selected spinal cord structures based on co-registration with anatomical spinal cord template for 25 subjects. One person data includes motion artifacts for which the proces failed. There are visible waves in sagittal images of some subjects caused probably by blood-vessel pulsing. Local quantification metrics of spinal cord anatomy (fractional anisotropy – FA, fractional volumes of first – f1 and second – f2 direction of anisotropic diffusion) from different parts (white matter, gray matter, cortico-spinal tract) and from different population groups (men vs. women), were extracted from dMRI data. As we expected, FA maps show visible decreases in areas of gray matter. We also detected second diffusion dirrection in slices, where the spinal roots come out. In some areas, fractional volume of second diffusion direction reaches up to 40% of the total component of the dMRI signal. All mentioned parameters probability density functions for all mentioned groups are non-normal distributions. Between male and female groups there were no significant distribution differences for f1 and f2 volumes. The distribution of FA values between men and women is statistically different. Unfortunatelly, there is a significant inter-subject variability in results, which has much higher dispersion than differences between different group distributions. Despite the inter-subject variability, this work significantly extends the knowledge about data acquisiton capabilities and MRI and dMRI data from cervical spinal cord image processing. This work also lays down foundations for utilization of the imaging method in future and planned clinical research, where it will be possible to test the alteration of the spinal cord anatomy on the minor secondary bundles separately.

Knowledge of the properties of single molecule diffusion is important for controlling dynamic self-assembly of molecular structures. A powerful experimental technique for determining diffusion coefficients is the recording of diffusion-induced signal fluctuations by a locally fixed point-like probe. Here, the signal becomes modified, whenever a molecule enters a certain detection area on the surface under the probe. The technique is minimal invasive and has a very good time resolution, enabling the investigation of highly mobile molecules. Theories are necessary for the analysis of the fluctuations and the extraction of diffusion properties. In this thesis, three methods are presented, which are based on the autocorrelation function, the distribution of peak widths and the distribution of interpeak intervals. Analytical expressions are derived for the distributions and the autocorrelation function in case of molecules, which can be described by circular or rectangular shapes. For rectangular shaped molecules, rotational diffusion can influence the recorded fluctuations. To allow for a simultaneous determination of rotational and translational diffusion coefficients the analytical treatment is extended. Furthermore, new methods are developed to determine the diffusion tensor for anisotropic stochastic molecular motion, using either one linearly extended probe or two individual probes. Coarse-graining the signal recorded by a point-like probe, which repeatedly moves on a line or a circle, is suggested for experimental implementation. All facets of the evaluation methods are verified against kinetic Monte Carlo simulations. Applications to experimental data, recorded by a locally fixed scanning tunneling microscope tip, are demonstrated for copperphthalocyanine and PTCDA molecules diffusing on Ag(100).

Diabetic Peripheral Neuropathy (DPN) is one of the most common types of diabetes that can affect the cornea. An accurate analysis of the corneal epithelium nerve structures and the corneal endothelial cell can assist early diagnosis of this disease and other corneal diseases, which can lead to visual impairment and then to blindness. In this thesis, fully-automated segmentation and quantification algorithms for processing and analysing sub-basal epithelium nerves and endothelial cells are proposed for early diagnosis of diabetic neuropathy in Corneal Confocal Microscopy (CCM) images. Firstly, a fully automatic nerve segmentation system for corneal confocal microscope images is proposed. The performance of the proposed system is evaluated against manually traced images with an execution time of the prototype is 13 seconds. Secondly, an automatic corneal nerve registration system is proposed. The main aim of this system is to produce a new informative corneal image that contains structural and functional information. Thirdly, an automated real-time system, termed the Corneal Endothelium Analysis System (CEAS) is developed and applied for the segmentation of endothelial cells in images of human cornea obtained by In Vivo CCM. The performance of the proposed CEAS system was tested against manually traced images with an execution time of only 6 seconds per image. Finally, the results obtained from all the proposed approaches have been evaluated and validated by an expert advisory board from two institutes, they are the Division of Medicine, Weill Cornell Medicine-Qatar, Doha, Qatar and the Manchester Royal Eye Hospital, Centre for Endocrinology and Diabetes, UK.

Les travaux de cette thèse portent sur des méthodes de volumes finis sur maillages quelconque pour la discrétisation de problèmes d'évolution non linéaires modélisant le transport de contaminants en milieu poreux et les écoulements diphasiques.Au Chapitre 1, nous étudions une famille de schémas numériques pour la discrétisation d'une équation parabolique dégénérée de convection-reaction-diffusion modélisant le transport de contaminants dans un milieu poreux qui peut être hétérogène et anisotrope. La discrétisation du terme de diffusion est basée sur une famille de méthodes qui regroupe les schémas de volumes finis hybrides, de différences finies mimétiques et de volumes finis mixtes. Le terme de convection est traité à l'aide d'une famille de méthodes qui s'appuient sur les inconnues hybrides associées aux interfaces du maillage. Cette famille contient à la fois les schémas centré et amont. Les schémas que nous étudions permettent une discrétisation localement conservative des termes d'ordre un et d'ordre deux sur des maillages arbitraires en dimensions d'espace deux et trois. Nous démontrons qu'il existe une solution unique du problème discret qui converge vers la solution du problème continu et nous présentons des résultats numériques en dimensions d'espace deux et trois, en nous appuyant sur des maillages adaptatifs.Au Chapitre 2, nous proposons un schéma de volumes finis hybrides pour la discrétisation d'un problème d'écoulement diphasique incompressible et immiscible en milieu poreux. On suppose que ce problème a la forme d'une équation parabolique dégénérée de convection-diffusion en saturation couplée à une équation uniformément elliptique en pression. On considère un schéma implicite en temps, où les flux diffusifs sont discrétisés par la méthode des volumes finis hybride, ce qui permet de pouvoir traiter le cas d'un tenseur de perméabilité anisotrope et hétérogène sur un maillage très général, et l'on s'appuie sur un schéma de Godunov pour la discrétisation des flux convectifs, qui peuvent être non monotones et discontinus par rapport aux variables spatiales. On démontre l'existence d'une solution discrète, dont une sous-suite converge vers une solution faible du problème continu. On présente finalement des cas test bidimensionnels.Le Chapitre 3 porte sur un problème d'écoulement diphasique, dans lequel la courbe de pression capillaire admet des discontinuité spatiales. Plus précisément on suppose que l'écoulement prend place dans deux régions du sol aux propriétés très différentes, et l'on suppose que la loi de pression capillaire est discontinue en espace à la frontière entre les deux régions, si bien que la saturation de l'huile et la pression globale sont discontinues à travers cette frontière avec des conditions de raccord non linéaires à l'interface. On discrétise le problème à l'aide d'un schéma, qui coïncide avec un schéma de volumes finis standard dans chacune des deux régions, et on démontre la convergence d'une solution approchée vers une solution faible du problème continu. Les test numériques présentés à la fin du chapitre montrent que le schéma permet de reproduire le phénomène de piégeage de la phase huile. / In Chapter 1 we study a family of finite volume schemes for the numerical solution of degenerate parabolic convection-reaction-diffusion equations modeling contaminant transport in porous media. The discretization of possibly anisotropic and heterogeneous diffusion terms is based upon a family of numerical schemes, which include the hybrid finite volume scheme, the mimetic finite difference scheme and the mixed finite volume scheme. One discretizes the convection term by means of a family of schemes which makes use of the discrete unknowns associated to the mesh interfaces, and contains as special cases an upwind scheme and a centered scheme. The numerical schemes which we study are locally conservative and allow computations on general multi-dimensional meshes. We prove that the unique discrete solution converges to the unique weak solution of the continuous problem. We also investigate the solvability of the linearized problem obtained during Newton iterations. Finally we present a number of numerical results in space dimensions two and three using nonconforming adaptive meshes and show experimental orders of convergence for upwind and centered discretizations of the convection term.In Chapter 2 we propose a finite volume method on general meshes for the numerical simulation of an incompressible and immiscible two-phase flow in porous media. We consider the case that it can be written as a coupled system involving a degenerate parabolic convection-diffusion equation for the saturation together with a uniformly elliptic equation for the global pressure. The numerical scheme, which is implicit in time, allows computations in the case of a heterogeneous and anisotropic permeability tensor. The convective fluxes, which are non monotone with respect to the unknown saturation and discontinuous with respect to the space variables, are discretized by means of a special Godunov scheme. We prove the existence of a discrete solution which converges, along a subsequence, to a solution of the continuous problem. We present a number of numerical results in space dimension two, which confirm the efficiency of the numerical method.Chapter 3 is devoted to the study of a two-phase flow problem in the case that the capillary pressure curve is discontinuous with respect to the space variable. More precisely we assume that the porous medium is composed of two different rocks, so that the capillary pressure is discontinuous across the interface between the rocks. As a consequence the oil saturation and the global pressure are discontinuous across the interface with nonlinear transmission conditions. We discretize the problem by means of a numerical scheme which reduces to a standard finite volume scheme in each sub-domain and prove the convergence of a sequence of approximate solutions towards a weak solution of the continuous problem. The numerical tests show that the scheme can reproduce the oil trapping phenomenon.

Atributos visuais de textura e forma foram investigados para a recuperação baseada em conteúdo de imagens mamográficas (CBIR). Para a similaridade de imagens, foi considerada a estrutura de densidade mamária, representada principalmente pelos tecidos fibro-glandulares. A pesquisa consistiu de três etapas: (1) Preparação e processamento das imagens; (2) Extração e seleção de atributos visuais de textura e forma; (3) Implementação de um sistema de recuperação de imagem. A primeira etapa consistiu dos processos de retirada de ruído do fundo da imagem, segmentação da região da mama, detecção da região de músculo peitoral, localização do mamilo e da segmentação da região de tecidos fibro-glandulares. Utilizou-se a equação de Difusão Anisotrópica com filtro de Wiener para retirada e suavização de ruídos encontrados na imagem e preservação da borda da mama. Para a segmentação da região da mama, foram utilizadas as técnicas de limiarização de Princípio de Máxima Entropia, Método de Preservação de Momento, Método de Otsu, Método interativo de Ridler & Carvard, Método de Reddi e Método da Matriz de Co-ocorrência. A melhor imagem foi escolhida numa tarefa supervisionada. A detecção automática da região do músculo peitoral foi feita com a combinação do operador de Canny e a transformada de Radon como detector de linha. A posição do mamilo foi detectada com a transformada de Radon como detector de direção de densidade. A segmentação da região de tecidos fibro-glandulares foi feita também com as técnicas de limiarização do Princípio de Máxima Entropia, Método de Preservação de Momento, e Método de Otsu. Momentos Estatísticos extraídos do Histograma, Medida de Granulometria, Momentos Estatísticos extraídos do Domínio de Radon, Momento de Hu, e Textura de Haralick foram investigados como atributos de textura. Medida de Área, Circularidade e Razão de Diâmetro foram investigados como atributos de forma. A rede de Mapas Auto-Organizáveis de Kohonen foi utilizada como sistema de recuperação de imagem. Foram utilizadas, neste trabalho, 1080 imagens do projeto de Banco de Imagens do HCFMRP-USP, módulo Mamografia. O treinamento e teste foram feitos com a técnica de \"leaving-one-out\" e os melhores resultados obtidos foram: Taxa de precisão de 91,07% para a combinação dos cinco grupos de atributos de Forma, Estatísticos Extraídos do Histograma, Momento de Hu, Espectral no Domínio de Radon e de Medida de Granulometria; taxa de precisão e revocação do coeficiente de correlação médio representadas pela área sob a curva com valor de 0,02351 dos grupos de atributos de forma, de Textura de Haralick e Momento de Hu. Os resultados obtidos indicaram a relevância de nosso trabalho e seu potencial de utilização para a recuperação baseada em conteúdo de imagens mamográficas. / Visual texture based on texture and shape features were investigated for content-based mammographic images retrieval (CBIR). For similarity of images, the mammary density structures were considered, mainly represented by fibro-glandular tissues. This research consisted of three stages: (1) Images preparation and processing; (2) Extraction and selection of the visual features; (3) Implementation of a retrieval system. The first stage consisted of noisy removing from the image background, breast region segmentation, pectoral muscle region detection, nipple localization and the fibro-glandular tissues region segmentation. The equation of Anisotropic Diffusion was used with Wiener filter for noisy removing with the breast region edge preservation. For the breast region segmentation, the Thresholding techniques were used of Maximum Entropy Principle, Moment Preserving Method, Otsu Method, Ridler & Carvard Method, Reddi Method and Co-occurrence Matrix Method. The better image was chosen in a supervised task. The automatic pectoral muscle region detection was made with the Canny operator and Radon Transform combination as straight line detector. The nipple position was detected with the Radon Transform as density direction detector. The fibro-glandular tissues region was also defined with the thresholding techniques of the Maximum Entropy Principle, Moment Preserving Method, and Otsu Method. The Statistical Moments extracted from the Histogram, Measured of Granulometry, Statistical Moments extracted in Radon Domain, Moment of Hu, and Haralick Textures were investigated as texture features. Area, Circularity and Diameter Ratio were investigated as shape features. The Self-Organizing Maps of Kohonen was used as image retrieval system. One thousand and eighty images of the HCFMRP-USP Database Project, Mammography Module, were used in this work. The training and test processes were realized with the \"leaving-one-out\" technique and the best results obtained were: The precision rate of 91,07% for the combination of the five following features group: Shape, Statistical Moments extracted of the Histogram, Moment of Hu, Statistical Moments extracted in Radon Domain and Measure of Granulometry; precision and revocation rates of the average coefficient of correlation represented by the area under the curve with value of 0,02351 for the three following features group: Shape, Haralick Textures and Moment de Hu. The results obtained indicated the relevance of our work for the content-based mammographic images retrieval.

Atributos visuais de textura e forma foram investigados para a recuperação baseada em conteúdo de imagens mamográficas (CBIR). Para a similaridade de imagens, foi considerada a estrutura de densidade mamária, representada principalmente pelos tecidos fibro-glandulares. A pesquisa consistiu de três etapas: (1) Preparação e processamento das imagens; (2) Extração e seleção de atributos visuais de textura e forma; (3) Implementação de um sistema de recuperação de imagem. A primeira etapa consistiu dos processos de retirada de ruído do fundo da imagem, segmentação da região da mama, detecção da região de músculo peitoral, localização do mamilo e da segmentação da região de tecidos fibro-glandulares. Utilizou-se a equação de Difusão Anisotrópica com filtro de Wiener para retirada e suavização de ruídos encontrados na imagem e preservação da borda da mama. Para a segmentação da região da mama, foram utilizadas as técnicas de limiarização de Princípio de Máxima Entropia, Método de Preservação de Momento, Método de Otsu, Método interativo de Ridler & Carvard, Método de Reddi e Método da Matriz de Co-ocorrência. A melhor imagem foi escolhida numa tarefa supervisionada. A detecção automática da região do músculo peitoral foi feita com a combinação do operador de Canny e a transformada de Radon como detector de linha. A posição do mamilo foi detectada com a transformada de Radon como detector de direção de densidade. A segmentação da região de tecidos fibro-glandulares foi feita também com as técnicas de limiarização do Princípio de Máxima Entropia, Método de Preservação de Momento, e Método de Otsu. Momentos Estatísticos extraídos do Histograma, Medida de Granulometria, Momentos Estatísticos extraídos do Domínio de Radon, Momento de Hu, e Textura de Haralick foram investigados como atributos de textura. Medida de Área, Circularidade e Razão de Diâmetro foram investigados como atributos de forma. A rede de Mapas Auto-Organizáveis de Kohonen foi utilizada como sistema de recuperação de imagem. Foram utilizadas, neste trabalho, 1080 imagens do projeto de Banco de Imagens do HCFMRP-USP, módulo Mamografia. O treinamento e teste foram feitos com a técnica de \"leaving-one-out\" e os melhores resultados obtidos foram: Taxa de precisão de 91,07% para a combinação dos cinco grupos de atributos de Forma, Estatísticos Extraídos do Histograma, Momento de Hu, Espectral no Domínio de Radon e de Medida de Granulometria; taxa de precisão e revocação do coeficiente de correlação médio representadas pela área sob a curva com valor de 0,02351 dos grupos de atributos de forma, de Textura de Haralick e Momento de Hu. Os resultados obtidos indicaram a relevância de nosso trabalho e seu potencial de utilização para a recuperação baseada em conteúdo de imagens mamográficas. / Visual texture based on texture and shape features were investigated for content-based mammographic images retrieval (CBIR). For similarity of images, the mammary density structures were considered, mainly represented by fibro-glandular tissues. This research consisted of three stages: (1) Images preparation and processing; (2) Extraction and selection of the visual features; (3) Implementation of a retrieval system. The first stage consisted of noisy removing from the image background, breast region segmentation, pectoral muscle region detection, nipple localization and the fibro-glandular tissues region segmentation. The equation of Anisotropic Diffusion was used with Wiener filter for noisy removing with the breast region edge preservation. For the breast region segmentation, the Thresholding techniques were used of Maximum Entropy Principle, Moment Preserving Method, Otsu Method, Ridler & Carvard Method, Reddi Method and Co-occurrence Matrix Method. The better image was chosen in a supervised task. The automatic pectoral muscle region detection was made with the Canny operator and Radon Transform combination as straight line detector. The nipple position was detected with the Radon Transform as density direction detector. The fibro-glandular tissues region was also defined with the thresholding techniques of the Maximum Entropy Principle, Moment Preserving Method, and Otsu Method. The Statistical Moments extracted from the Histogram, Measured of Granulometry, Statistical Moments extracted in Radon Domain, Moment of Hu, and Haralick Textures were investigated as texture features. Area, Circularity and Diameter Ratio were investigated as shape features. The Self-Organizing Maps of Kohonen was used as image retrieval system. One thousand and eighty images of the HCFMRP-USP Database Project, Mammography Module, were used in this work. The training and test processes were realized with the \"leaving-one-out\" technique and the best results obtained were: The precision rate of 91,07% for the combination of the five following features group: Shape, Statistical Moments extracted of the Histogram, Moment of Hu, Statistical Moments extracted in Radon Domain and Measure of Granulometry; precision and revocation rates of the average coefficient of correlation represented by the area under the curve with value of 0,02351 for the three following features group: Shape, Haralick Textures and Moment de Hu. The results obtained indicated the relevance of our work for the content-based mammographic images retrieval.