Este trabalho apresenta uma estratégia computacional para a consolidação do treinamento dos radiologistas residentes por meio da classificação de imagens mamográficas pela similaridade, analisando informações dos laudos realizados por médicos experientes, obtendo os atributos extraídos das imagens médicas. Para a descoberta de padrões que caracterizam a similaridade aplicam-se técnicas de processamento digital de imagens e de mineração de dados nas imagens mamográficas. O reconhecimento de padrões tem como objetivo realizar a classificação de determinados conjuntos de imagens em classes. A classificação dos achados mamográficos é realizada utilizando Redes Neurais Artificiais, por meio do classificador Self-Organizing Map (SOM). O presente trabalho utiliza a recuperação de imagens por conteúdo (CBIR- Content-Based Image Retrieval), considerando a similaridade em relação a uma imagem previamente selecionada para o treinamento. As imagens são classificadas de acordo com a similaridade, analisando-se informações dos atributos extraídos das imagens e dos laudos. A identificação da similaridade é obtida pela extração de características, com a utilização da transformada de wavelets. / This work presents a computational strategy to consolidate the training of residents radiologists through the classification of mammographic images by similarity, analyzing information from reports made by experienced physicians, obtaining the attributes extracted from medical images. For the discovery of patterns that characterize the similarity apply techniques of digital image processing and data mining in mammographic images. Pattern recognition aims to achieve the classification of certain sets of images in classes. The classification of mammographic is performed using Artificial Neural Networks, through the classifier Self-Organizing Map (SOM). This work uses the image retrieval (CBIR-Content- Based Image Retrieval), considering the similarity in relation to an image already selected for training. The images are classified according to similarity, analyzing attribute information extracted from the images and reports. The identification of similarity was obtained by feature extraction, using the technique of wavelet transform.

As Redes Neurais Artificiais (RNA) são modelos matemáticos associados à inteligência computacional capaz de aprender e generalizar informações, podendo assim ser utilizada como um classificador de imagens. O presente trabalho objetiva analisar o espelho nasal bovino com o intuito de comprovar que é uma característica única e permanente do animal podendo assim, ser sua identificação única. O experimento foi dividido em duas etapas. Para compor o banco de dados da primeira etapa foram utilizados 51 bovinos da raça Nelore com idade média de 11 meses, dos quais foram coletadas para a formação do banco de dados dezesseis imagens de cada animal, totalizando uma base de 816 imagens. Na segunda etapa do experimento foram utilizados 16 bovinos do banco de dados inicial, escolhidos de forma aleatória, com idade média de 23 meses. Destes foram coletadas 11 imagens para verificar se os padrões do espelho nasal, com o passar dos meses, mantêm seu padrão tornando possível, assim, a identificação do animal. Os algoritmos de processamento digital de imagens foram implementados utilizando o software MATLAB®. Após o processamento das imagens, as características vetorizadas foram utilizadas para treinamento e teste de uma rede neural artificial utilizando o algoritmo MLP, implementado usando o compilador C DGW, que serviu como classificador das mesmas. Também foi utilizado o algoritmo do K vizinhos mais próximos (K-nn), para realizar os testes de classificação, usando um método estatistico. A validação do classificador foi realizada mediante análise estatística dos seus erros e acertos. O erro médio quadrático utilizado neste estudo foi menor que 1%. Os resultados apresentados pelo classificador K-nn foram maiores que o da Rede Neural Artificial, porém ambos não alcançaram acertos acima de 90%, o que é considerado adequado a um classificador. Pode-se concluir que o método utilizado para extração de características não apresentou uma boa representatividade, porém ainda assim foi possível observar a tendência classificatória dos animais através das características do espelho nasal, assim como a tendência da permanência dos padrões com o envelhecimento do animal. / Artificial Neural Networks (ANN) are mathematical models associated with artificial intelligence that can learn and generalize information, therefore they can be used as images classifiers. This paper aims to analyze the cattle muzzle in order to prove that it is a unique and permanent characteristic of the animal thus, being used as its unique identification. The experiment was divided into two stages. To make the database of the first phase were used 51 Nelore bovines with an average age of 11 months, from which sixteen images of each animal were collected totalling of 816 images for the database. In the second stage of the experiment 16 bovines from the initial database were used, chosen randomly, with an average age of 23 months. From those 11 images were collected to verify if the standards of the muzzle remain the same after a couple of months, so the animal can be identified. The processing digital image algorithms were implemented using MATLAB® software. After the images processing, vectorized features were used to train and test an artificial neural network using the MLP algorithm, implemented using the C compiler DGW, and was used as a classifier. We also used the algorithm of K nearest neighbors (Knn) to perform the classification tests using a statistical method. The validation of the classifier was performed using statistical analysis of their mistakes and successes. The average square error used in this study was less than 1%. The results presented by K-nn classifier were higher than the one of Artificial Neural Network; nevertheless, both failed to reach above 90% success, which is considered suitable for a classifier. It can be concluded that the method used for feature extraction did not show a good performance, although it was possible to observe the trend of classification of animals through the characteristics of the muzzle, as well as the tendency of the permanence of the standards with the animal aging.

This study is part of the Diavik Waste-Rock Pile Project taking place at the Diavik Diamond Mine in the Northwest Territories, Canada. The project involves the construction of three 15m-scale low sulfide test waste-rock piles and monitoring of fluid flow, geochemical reactions, heat and gas transport within the waste-rock piles and characterization of the physical properties of the waste-rock piles. The focus of this thesis is characterizing grain-size distribution of the waste-rock and quantifying gas transport in the test waste-rock piles. Grain size of waste rock ranges from millimeters to meters. Sieve analysis typically only provides information of grain size <0.1 m at a single location. A computer program was developed using digital image-processing techniques to obtain a spatial grain-size distribution from photographs of tip faces of the test waste-rock piles acquired in the field. The program characterizes grain size >0.1 m and employs a region-growing algorithm for segmentation of waste-rock grains with pre- and post-processing techniques to improve the accuracy of segmentation. The program was applied to photographs of six different tip faces of the test waste-rock piles. For grain size <0.1 m, data from sieve analyses were attached to the grain-size curves generated from image grain-size analyses to obtain a full spectrum grain-size analyses ranging from boulders to fines. The results show that fine fractions are retained at the top of the tip faces and grain size increases non-linearly from top to bottom of a waste-rock pile. Calculations show that although the greatest mass is associated with the medium and coarse fractions, the greatest surface area is associated with the fine fractions. The results are consistent with field observation that the initial solute concentrations are greatest at the top of the pile and saturated hydraulic conductivity are lower at the top of the pile than in the pile interior. Statistical moments show that the test waste-rock piles have mean grain size of granules and are very poorly sorted, coarse skewed and leptokurtic. Permeability is calculated using empirical formulae and good agreement is obtained between calculated values and field measurements. The heterogeneity of grain size obtained from this study can provide a basis for future modeling efforts. Gas transport analysis focused on 1) substantiating the relationship between wind flow external to the waste-rock pile and gas pressures within the pile, 2) determining the gas flow regime in the pile, and 3) quantifying the temporal variation in wind speed and direction and determining the relevant time scales. Differential gas pressures were measured in 2008 at 49 locations within one of the three test waste-rock piles and 14 locations on the surface of the pile at one-minute intervals. Wind speed and direction were measured at 10-min intervals. Correlations between wind vectors and pressure measurements show that the wind influences pressure fluctuations in the test pile. The strength of the correlation is roughly inversely proportional to the distance between measurement ports and the atmospheric boundary. The linear relationship between internal pressure measurements and surface pressure measurements demonstrate that gas flow is Darcian within the test waste-rock pile. Spectral analysis of wind data and a one-dimensional analytical solution to the flow equations show that the persistence of wind in a certain direction has most pronounced effects on transient gas flow within the pile. The penetration depth of wind-induced gas pressure wave is a function of the periodicity of the wind and permeability of the waste-rock pile.

Inspection of LCD light-guide plate using digital image processing is proposed. Binary dot-pattern images from SEM observation are obtained by image segmentation. Pattern recognition for the images is then performed using moment invariants, Bayes classifier, and Neural network. The rotation independent classification for the recognition using only one descript shape factor are also proposed to reduce storage space. It is found the method has been applied successfully in inspection of different defects on the plate subject to any rotation angles and image scales.

Fractal was first used in measuring the length of the coastline, with the fractal research and development, not only to break the traditional Archimedean geometry, but also to explain many scientific to ignore the complexity and nature of nonlinear phenomena structure .Fractal has been widely applied to such as physics, astronomy, geography and sociology and other fields, as a wave of interdisciplinary research in recent years. Coastal areas has always been cultural, economic and activities areas since ancient times. Coastal zone was land and sea for the interaction region by a variety of factors (ex: waves, tides, currents and wind, etc.) continue to function, derived from different coastal terrain. Therefore changes in the coast of the deep impact of humanity. Under the principle of the conservation and development, Coastal areas should be use of modern technology to prediction, analysis, assessment, planning, and management, so that a sustainable preservation of coastal resources. In this study, static and dynamic predict and simulation the coast shape base on fractal. The static part is observation of 29 beaches in South China coast. And collect and calculate the parameters and fractal dimensions of the coast. Through the shape of image processing and analysis of information, to find two generators of the coast. Through the data mining technology to identify the criteria for classification, and to simulation the coastline by generate iterations method. The dynamic part is based on hydraulic model¡¦s results, the use of traditional multiple linear regression and neural network to compare the dynamic prediction of the coastline. The results show that the use of neural networks to predict than the use of multiple linear regression, and effect of use difference angle (£c) to predict sub-coastlines than the effect of not use difference angle (£c) to predict, and add fractal dimension can effectively reduce the predict error and increase the degree of interpretation.

As more and more multimedia applications are integrated in mobile devices, a significant amount of energy is devoted to digital signal processing (DSP). Thus, reducing energy consumption for DSP systems has become an important design goal for battery operated mobile devices. Since supply voltage scaling is one of the most effective methods to reduce power/energy consumption, this study examines aggressive voltage scaling to achieve significant energy savings by allowing some output quality degradation for error tolerant image processing system. The objective of proposed research is to explore ultra-low energy image processing system design methodologies based on efficient accuracy (quality)-energy tradeoffs. This dissertation presents several new analyses and techniques to achieve significant energy savings without noticeable quality degradation under aggressive voltage scaling. In the first, this work starts from accurate error analysis and a model based on input sequence dependent delay estimation. Based on the analysis, we explain the dependence of voltage scalability on input image types, which may be used for input dependent adaptive control for optimal accuracy-energy tradeoffs. In addition, this work includes the system-level analysis of the impact of aggressive voltage scaling on overall energy consumption and a low-cost technique to reduce overall energy consumption. Lastly, this research exploits an error concealment technique to improve the efficiency of accuracy-energy tradeoffs. For an image compression system, the technique minimizes the impact of delay errors on output quality while allowing very low voltage operations for significant energy reduction.

This study is part of the Diavik Waste-Rock Pile Project taking place at the Diavik Diamond Mine in the Northwest Territories, Canada. The project involves the construction of three 15m-scale low sulfide test waste-rock piles and monitoring of fluid flow, geochemical reactions, heat and gas transport within the waste-rock piles and characterization of the physical properties of the waste-rock piles. The focus of this thesis is characterizing grain-size distribution of the waste-rock and quantifying gas transport in the test waste-rock piles. Grain size of waste rock ranges from millimeters to meters. Sieve analysis typically only provides information of grain size <0.1 m at a single location. A computer program was developed using digital image-processing techniques to obtain a spatial grain-size distribution from photographs of tip faces of the test waste-rock piles acquired in the field. The program characterizes grain size >0.1 m and employs a region-growing algorithm for segmentation of waste-rock grains with pre- and post-processing techniques to improve the accuracy of segmentation. The program was applied to photographs of six different tip faces of the test waste-rock piles. For grain size <0.1 m, data from sieve analyses were attached to the grain-size curves generated from image grain-size analyses to obtain a full spectrum grain-size analyses ranging from boulders to fines. The results show that fine fractions are retained at the top of the tip faces and grain size increases non-linearly from top to bottom of a waste-rock pile. Calculations show that although the greatest mass is associated with the medium and coarse fractions, the greatest surface area is associated with the fine fractions. The results are consistent with field observation that the initial solute concentrations are greatest at the top of the pile and saturated hydraulic conductivity are lower at the top of the pile than in the pile interior. Statistical moments show that the test waste-rock piles have mean grain size of granules and are very poorly sorted, coarse skewed and leptokurtic. Permeability is calculated using empirical formulae and good agreement is obtained between calculated values and field measurements. The heterogeneity of grain size obtained from this study can provide a basis for future modeling efforts. Gas transport analysis focused on 1) substantiating the relationship between wind flow external to the waste-rock pile and gas pressures within the pile, 2) determining the gas flow regime in the pile, and 3) quantifying the temporal variation in wind speed and direction and determining the relevant time scales. Differential gas pressures were measured in 2008 at 49 locations within one of the three test waste-rock piles and 14 locations on the surface of the pile at one-minute intervals. Wind speed and direction were measured at 10-min intervals. Correlations between wind vectors and pressure measurements show that the wind influences pressure fluctuations in the test pile. The strength of the correlation is roughly inversely proportional to the distance between measurement ports and the atmospheric boundary. The linear relationship between internal pressure measurements and surface pressure measurements demonstrate that gas flow is Darcian within the test waste-rock pile. Spectral analysis of wind data and a one-dimensional analytical solution to the flow equations show that the persistence of wind in a certain direction has most pronounced effects on transient gas flow within the pile. The penetration depth of wind-induced gas pressure wave is a function of the periodicity of the wind and permeability of the waste-rock pile.

Η παρούσα διπλωματική εργασία ασχολείται με την εκτίμηση του χρώματος αντικειμένων τα οποία απεικονίζονται σε φωτογραφίες, μέσω χρωματικής ισοστάθμισης και κατά συνέπεια μέσω διαχωρισμού των συνιστωσών ανάκλασης του φωτός. Συγκεκριμένα, εφαρμόσθηκε μια μέθοδος εκτίμησης του χρώματος αντικειμένων χρησιμοποιώντας μια βάση δεδομένων πολλαπλών εικόνων οι οποίες επεξεργάστηκαν ψηφιακά. Οι φωτογραφίες της βάσης δεδομένων απεικονίζουν την ίδια σκηνή και λήφθηκαν τοποθετώντας τη φωτεινή πηγή σε διαφορετική θέση στην περίπτωση κάθε φωτογραφίας. Η εργασία βασίστηκε κατά ένα μεγάλο μέρος σε άρθρα και πληροφορίες που συλλέχθηκαν από το διαδίκτυο. Τα πειραματικά αποτελέσματα ήταν ικανοποιητικά και έδειξαν πως ο συγκεκριμένος αλγόριθμος επιτυγχάνει χρωματική ισοστάθμιση και απόδοση χρωμάτων με καλή ακρίβεια και μικρή υπολογιστική πολυπλοκότητα. / The present Diploma Thesis deals with color estimation of objects which are depicted in photos, through color constancy and thus by separating the reflection components of light. We applied a method for estimating the color of objects using a database of multiple images that were digitally processed. The photos of the database represent the same scene and were obtained by placing the light source in a different location while capturing each photo. The present work was largely based on articles and information gathered from the internet. The experimental results were satisfactory and showed that the specific algorithm achieves color constancy and color estimation with good accuracy and low computational complexity.

Made available in DSpace on 2014-06-11T19:23:31Z (GMT). No. of bitstreams: 0 Previous issue date: 2004Bitstream added on 2014-06-13T20:10:58Z : No. of bitstreams: 1 zanin_rb_me_prud.pdf: 6131122 bytes, checksum: c6c482e83836f2b30cf13b6d9d31f02a (MD5) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Pesquisas em extração de rodovias em imagens digitais não são recentes, sendo as primeiras da década de 70. Os métodos desenvolvidos normalmente são classificados em duas classes: semi-automático e automático. Esta última classe não prevê a intervenção de um operador. Neste contexto, a extração automática de cruzamentos de rodovias é fundamental, embora muito poucos trabalhos são encontrados na literatura sobre extração de rodovias. A razão principal é a grande diversidade de cruzamentos de rodovias, gerando algumas dificuldades para construir um conhecimento a priori sobre cruzamentos. Esta pesquisa propõe uma metodologia automática para resolver este problema, combinando segmentos de rodovias extraídos de uma imagem de alta resolução (pixel < 0,7 m), com linhas extraídas numa imagem de baixa resolução (pixel > 2m), reamostrada da imagem original de alta resolução. A metodologia proposta foi testada em um conjunto de imagens de alta resolução, mostrando-se confiável e eficiente. Desde que as bordas das rodovias sejam satisfatoriamente definidas, o método se mostrou capaz de extrair totalmente os cruzamentos de rodovias. Além disso, gerou uma significante melhora na malha viária (aproximadamente 16%) extraída pela metodologia automática de extração de segmentos de rodovias. / Researches on road extraction from digital images are not recent, being the first one from 70's. The methods developed are usually classified into two classes, i.e., semiautomatic and automatic. Concerning this last class, no intervention with the operator is expected. In this context, the road crossing extraction is fundamental, although very few works are found in the relevant literature. The main reason is the great diversity of road crossings, bringing some difficulties to build up a priori knowledge of them. This research proposes a methodology for solving this problem combining road segments extracted from a high – resolution image (pixel < 0.7 m), with lines extracted from a low - resolution image (pixel > 2m) resampled from the original, high – resolution image. The proposed methodology was tested with a set of high – resolution image, showing that it is reliable and efficient. Whenever the road edges were well - defined the method was able to totally extract the road crossings. In addition it provided road networks with completion significantly better (about 16%) than the corresponding ones previously extracted by the automatic road segment extraction method.

Apesar dos avanços recentes na tecnologia dos aparelhos de ressonância magnética (RM) permitirem a aquisição de imagens de alta resolução, ainda não é possível delinear de forma confiável os limites entre regiões de diferentes citoarquiteturas baseando-se somente nesta modalidade. As imagens de histologia são mandatórias quando se necessita saber o limite exato entre diferentes regiões neuroanatômicas. Contudo, o processamento histológico inevitavelmente causa grandes deforma¸coes no tecido, o que torna a compara- ção direta entre as duas modalidades inviável. Os estudos de neuroimagem/neuroanatomia que necessitam de comparação com a histologia devem necessariamente incluir uma etapa de alinhamento entre as duas modalidades; tarefa que muitas vezes acaba sendo realizada manualmente. Entretanto, o registro manual ´e demorado e pouco acurado, se tornando inviável quando os exames de histologia geram centenas de imagens. Este trabalho propõe um método para registro de imagens de histologia e RM, composto por um conjunto de recomenda¸coes para o preparo das imagens cujo objetivo ´e otimizá-las para o registro; e por uma pipeline computacional capaz de registrar as imagens consideradas. O trabalho aqui descrito foi desenvolvido primeiramente com o intuito de registrar imagens de espécimens de hipocampo provenientes do projeto CINAPCE e, posteriormente, para registro de imagens de encéfalo inteiro provenientes do Banco de Cérebros da Faculdade de Medicina da Universidade de São Paulo. A pipeline computacional foi testada com sucesso em imagens reais de dois encéfalos inteiros. A avaliação quantitativa dos registros realizados foi feita comparando segmenta¸coes manuais do hipocampo direito, núcleo caudado esquerdo e ventrículos laterais superiores, realizadas no volume de RM e da histologia registrada. A quantificação do resultado foi feita através do cálculo das métricas coeficiente de Dice (CSD) e distancia espectral ponderada (DEP) sobre as segmentações. A pipeline obteve um CSD médio de aproximadamente 0,77 e um DEP médio de aproximadamente 0,003. Os resultados mostraram que o método foi capaz de registrar as imagens de histologia nas respectivas imagens de RM exigindo interação mínima com o usuário. / Although latest advances in MRI technology have allowed the acquisition of higher resolution images, reliable delineation of cytoarchitectural boundaries is not yet possible based solely on that modality. Histological images are regularly required to locate the exact limits between neuroanatomical structures. Histological processing is nevertheless prone to cause a high amount of tissue distortion, which prevents direct comparison between the two modalities. Neuroimage/neuroanatomy studies that require direct comparison between histology am MRI must include a registration step. Such task is usually manually performed, but that becames infeasible for large histology volumes. Moreover, manual registration is time consuming and inaccurate. This thesis proposes a set of tissue processing recommendations aiming at optimizing the registration proccess, together with a computational pipeline for registering histology to MRI. The herein described work was initially designed to proccess hippocampi specimens from the CINAPCE project and posteriorly improved to process full brain images from the Brain Bank of the Brazilian Aging Brain Study Group. The pipeline was tested on two full brain histology volumes from the Brain Bank of the Brazilian Aging Brain Study Group. Results were assessed by comparison of manual segmentations of the left caudate nucleus, right hippocampus and superior lateral ventricles, performend on both MRIs and registered histology volumes. Quatitative evaluation was performed by computing the Dice coeficient (DC) and normalized weighted spectral distance (WESD) on the segmentations. The pipeline precessing yielded mean DC of 0.77 and mean WESD of 0.0033. The described method was able to sucessfuly register histology to their corresponding MRI volumes with minimal user interaction.