La segmentation d’images est une opération cruciale pour le traitement d’images. Elle est toujours le point de départ des processus d’analyse de formes, de détection de mouvement, de visualisation, des estimations quantitatives de distances linéaires, de surfaces et de volumes. À ces fins, la segmentation consiste à catégoriser les voxels en des classes basées sur leurs intensités locales, leur localisation spatiale et leurs caractéristiques de forme ou de voisinage. La difficulté de la stabilité des résultats des méthodes de segmentation pour les images médicales provient des différents types de bruit présents.Dans ces images, le bruit prend deux formes : un bruit physique dû au système d’acquisition, dans notre cas l’IRM (Imagerie par Résonance Magnétique), et le bruit physiologique dû au patient. Ces bruits doivent être pris en compte pour toutes les méthodes de segmentation d’images. Durant cette thèse,nous nous sommes focalisés sur des modèles Multi-Agents basés sur les comportements biologiques des araignées et des fourmis pour effectuer la tâche de segmentation. Pour les araignées, nous avons proposé une approche semi-automatique utilisant l’histogramme de l’image pour déterminer le nombre d’objets à détecter. Tandis que pour les fourmis, nous avons proposé deux approches : la première dite classique qui utilise le gradient de l’image et la deuxième, plus originale, qui utilise une partition intervoxel de l’image. Nous avons également proposé un moyen pour accélérer le processus de segmentation grâce à l’utilisation des GPU (Graphics Processing Unit). Finalement, ces deux méthodes ont été évaluées sur des images d’IRM de cerveau et elles ont été comparées aux méthodes classiques de segmentation : croissance de régions et Otsu pour le modèle des araignées et le gradientde Sobel pour les fourmis. / Image segmentation is a crucial operation for image processing. It is always the starting point of shape analysis process, motion detection, visualization, and quantitative estimation of linear distances, surfaces and volumes. For this, the segmentation consists on classifying the voxels into classes based on their local strengths, their spatial location and shape characteristics or neighborhood. The difficulty of the results stability of segmentation methods for medical images comes from the different types of noise present inside every image. In these images, the noise takes two forms: a physical noise due to the acquisition system, in our case, MRI (Magnetic Resonance Imaging), and a physiological noise due to the patient. These noises should be considered for all methods of segmentation. In this thesis, we focused on Multi-Agent models based on the biological behavior of spiders and ants to perform the task of segmentation. For spiders, we proposed a semi-automatic method using the histogram of the image to determine the number of objects to be detected. As for ants, we proposed two approaches: one that uses the so-called classical gradient of the image and the second, more original, which uses an intervoxel partition of the image. We also proposed a way to speed up the segmentation process through the use of the GPU (Graphics Processing Unit). Finally, these two methods were evaluated on MR images of brain and were compared with conventional methods of segmentation: region growing and Otsu for the model of spiders and Sobel gradient for the ants.

A possibilidade de armazenamento de imagens no formato digital favoreceu a evolução de diversos ramos de atividades, especialmente as áreas de pesquisa e clínica médica. Ao mesmo tempo, o volume crescente de imagens armazenadas deu origem a um problema de relevância e complexidade consideráveis: a Recuperação de Imagens Baseada em Conteúdo, que, em outras palavras, diz respeito à capacidade de um sistema de armazenamento processar operações de consulta de imagens a partir de características visuais, extraídas automaticamente por meio de métodos computacionais. Das principais questões que constituem este problema, amplamente conhecido pelo termo CBIR - Content-Based Image Retrieval, fazem parte as seguintes: Como interpretar ou representar matematicamente o conteúdo de uma imagem? Quais medidas que podem caracterizar adequadamente este conteúdo? Como recuperar imagens de um grande repositório utilizando o conteúdo extraído? Como estabelecer um critério matemático de similaridade entre estas imagens? O trabalho desenvolvido e apresentado nesta tese busca, exatamente, responder perguntas deste tipo, especialmente para os domínios de imagens médicas e da biologia genética, onde a demanda por sistemas computacionais que incorporam técnicas CBIR é consideravelmente alta por diversos motivos. Motivos que vão desde a necessidade de se buscar informação visual que estava até então inacessível pela falta de anotações textuais, até o interesse em poder contar com auxílio computacional confiável para a importante tarefa de diagnóstico clínico. Neste trabalho são propostos métodos e soluções inovadoras para o problema de segmentação e extração de características de imagens médicas e imagens de padrões espaciais de expressão genética. A segmentação é o processo de delimitação automático de regiões de interesse da imagem que possibilita uma caracterização bem mais coerente do conteúdo visual, comparado com as tradicionais técnicas de caracterização global e direta da imagem. Partindo desta idéia, as técnicas de extração de características desenvolvidas neste trabalho empregam métodos adaptativos de segmentação de imagens e alcançam resultados excelentes na tarefa de recuperação baseada em conteúdo / Storing images in digital format has supported the evolution of several branches of activities, specially the research area and medical clinic. At the same time, the increasing volume of stored images has originated a topic of considerable relevance and complexity: the Content- Based Imagem Retrieval, which, in other works, is related to the ability of a computational system in processing image queries based on visual features automatically extracted by computational methods. Among the main questions that constitute this issue, widely known as CBIR, are these: How to mathematically express image content? What measures can suitably characterize this content? How to retrieve images from a large dataset employing the extracted content? How to establish a mathematical criterion of similarity among the imagens? The work developed and presented in this thesis aims at answering questions like those, especially for the medical images domain and genetical biology, where the demand for computational systems that embody CBIR techniques is considerably high for several reasons. Reasons that range from the need for retrieving visual information that was until then inaccessible due to the lack of textual annotations, until the interest in having liable computational support for the important task of clinical diagnosis. In this work are proposed innovative methods and solutions for the problem of image segmentation and feature extraction of medical images and images of gene expression patterns. Segmentation is the process that enables a more coherent representation of image?s visual content than that provided by traditional methods of global and direct representation. Grounded in such idea, the feature extraction techniques developed in this work employ adaptive image segmentation methods, and achieve excellent results on the task of Content-Based Image Retrieval

Uma das etapas mais importantes da análise de imagens e, que conta com uma enorme quantidade de aplicações, é a segmentação. No entanto, uma boa parte das técnicas tradicionais apresenta alto custo computacional, dificultando sua aplicação em imagens de alta resolução como, por exemplo, as imagens de ninhais de aves do Pantanal que também serão analisadas neste trabalho. Diante disso, é proposta uma nova abordagem de segmentação que combina algoritmos de detecção de comunidades, pertencentes à teoria das redes complexas, com técnicas de extração de superpixels. Tal abordagem é capaz de segmentar imagens de alta resolução mantendo o compromisso entre acurácia e tempo de processamento. Além disso, como as imagens de ninhais analisadas apresentam características peculiares que podem ser mais bem tratadas por técnicas de segmentação por textura, a técnica baseada em Markov Random Fields (MRF) é proposta, como um complemento à abordagem de segmentação inicial, para realizar a identificação final das aves. Por fim, devido à importância de avaliar quantitativamente a qualidade das segmentações obtidas, um nova métrica de avaliação baseada em ground-truth foi desenvolvida, sendo de grande importância para a área. Este trabalho contribuiu para o avanço do estado da arte das técnicas de segmentação de imagens de alta resolução, aprimorando e desenvolvendo métodos baseados na combinação de redes complexas com superpixels, os quais alcançaram resultados satisfatórios com baixo tempo de processamento. Além disso, uma importante contribuição referente ao censo demográfico de aves por meio da análise de imagens aéreas de ninhais foi viabilizada por meio da aplicação da técnica de segmentação MRF. / Segmentation is one of the most important steps in image analysis with a large range of applications. However, some traditional techniques exhibit high computational costs, hindering their application in high resolution images such as the images of birds nests from Pantanal, one of Brazilian most important wetlands. Therefore, we propose a new segmentation approach that combines community detection algorithms, originated from the theory of the complex networks, with superpixels extraction techniques. This approach is capable of segmenting high resolution images while maintaining the trade-off between accuracy and processing time. Moreover, as the nest images exhibit peculiar characteristics that can be better dealt with texture segmentation techniques, the Markov Random Fields (MRF) technique is proposed, as a complement to the initial approach, to perform the final identification of the birds. Finally, due to the importance of the quantitatively evaluation of the segmentation quality, a new evaluation metric based on ground-truth was developed, being of great importance to the segmentation field. This work contributed to the state of art of high resolution images segmentation techniques, improving and developing methods based on combination of complex networks and superpixels, which generated satisfactory results within low processing time. Moreover, an important contribution for the birds census by the analysis of aerial images of birds nests was made possible by application of the MRF technique.

Plâncton são organismos microscópicos que constituem a base da cadeia alimentar de ecossistemas aquáticos. Eles têm importante papel no ciclo do carbono pois são os responsáveis pela absorção do carbono na superfície dos oceanos. Detectar, estimar e monitorar a distribuição das diferentes espécies são atividades importantes para se compreender o papel do plâncton e as consequências decorrentes de alterações em seu ambiente. Parte dos estudos deste tipo é baseada no uso de técnicas de imageamento de volumes de água. Devido à grande quantidade de imagens que são geradas, métodos computacionais para auxiliar no processo de análise das imagens estão sob demanda. Neste trabalho abordamos o problema de identificação da espécie. Adotamos o pipeline convencional que consiste dos passos de detecção de alvo, segmentação (delineação de contorno), extração de características, e classificação. Na primeira parte deste trabalho abordamos o problema de escolha de um algoritmo de segmentação adequado. Uma vez que a avaliação de resultados de segmentação é subjetiva e demorada, propomos um método para avaliar algoritmos de segmentação por meio da avaliação da classificação no final do pipeline. Experimentos com esse método mostraram que algoritmos de segmentação distintos podem ser adequados para a identificação de espécies de classes distintas. Portanto, na segunda parte do trabalho propomos um método de classificação que leva em consideração múltiplas segmentações. Especificamente, múltiplas segmentações são calculadas e classificadores são treinados individualmente para cada segmentação, os quais são então combinados para construir o classificador final. Resultados experimentais mostram que a acurácia obtida com a combinação de classificadores é superior em mais de 2% à acurácia obtida com classificadores usando uma segmentação fixa. Os métodos propostos podem ser úteis para a construção de sistemas de identificação de plâncton que sejam capazes de se ajustar rapidamente às mudanças nas características das imagens. / Plankton are microscopic organisms that constitute the basis of the food chain of aquatic ecosystems. They have an important role in the carbon cycle as they are responsible for the absorption of carbon in the ocean surfaces. Detecting, estimating and monitoring the distribution of plankton species are important activities for understanding the role of plankton and the consequences of changes in their environment. Part of these type of studies is based on the analysis of water volumes by means of imaging techniques. Due to the large quantity of generated images, computational methods for helping the process of image analysis are in demand. In this work we address the problem of species identification. We follow the conventional pipeline consisting of target detection, segmentation (contour delineation), feature extraction, and classification steps. In the first part of this work we address the problem of choosing an appropriate segmentation algorithm. Since evaluating segmentation results is a subjective and time consuming task, we propose a method to evaluate segmentation algorithms by evaluating the classification results at the end of the pipeline. Experiments with this method showed that distinct segmentation algorithms might be appropriate for identifying species of distinct classes. Therefore, in the second part of this work we propose a classification method that takes into consideration multiple segmentations. Specifically, multiple segmentations are computed and classifiers are trained individually for each segmentation, which are then combined to build the final classifier. Experimental results show that the accuracy obtained with the combined classifier is superior in more than 2% to the accuracy obtained with classifiers using a fixed segmentation. The proposed methods can be useful to build plankton identification systems that are able to quickly adjust to changes in the characteristics of the images.

In the scheme of Non Destructive Testing (NDT), defect detection is an important process. Traditional image processing techniques have successfully been used for defect recognition. Usage of machine learning techniques is still in the initial stages of development. Convolution Neural Networks (CNN) is widely used for object classification one such scenario is defect classification in weld tubes. With the advent of deep learning techniques such as transfer learning, we can transfer knowledge gained in one domain successfully into other. Pre-trained models successfully learn features from large scale datasets that can be used for in domains having sparse data and smaller datasets. The aim of this work is to help a manual inspector in recognition of defects on the weld tubes. With a given set of images, we proceed by forming unique pipeline architecture for automatic defect recognition. The research in this thesis focuses on extraction of welds using image segmentation techniques, creating a dataset of defects and using it to on pre-trained Convolution Neural Networks of VGG16, VGG19, Inception V3 and ResNet101. We evaluate the models on different metrics finding the best suited model for the created dataset. Further a prototype sliding window solution is used to find defects over the extracted weld region. We also present the limitations of this approach and suggest modifications that can be implemented in the future.

This work examines the use of convolutional neural networks with a focus on semantic and instance segmentation of cells from microscopic images. The theoretical part contains a description of deep neural networks and a summary of widely used convolutional architectures for image segmentation. The practical part of the work is devoted to the creation of a convolutional neural network model based on the U-Net architecture. It also contains cell segmentation of predicted images using three methods, namely thresholding, the watershed and the random walker.

<div>U-net is an image segmentation technique developed primarily for medical image analysis that can precisely segment images using a scarce amount of training data. These traits provide U-net with a high utility within the medical imaging community and have resulted in extensive adoption of U-net as the primary tool for segmentation tasks in medical imaging. The success of U-net is evident in its widespread use in nearly all major image modalities from CT scans and MRI to X-rays and microscopy. Furthermore, while U-net is largely a segmentation tool, there have been instances of the use of U-net in other applications. Given that U-net's potential is still increasing, this review examines the numerous developments and breakthroughs in the U-net architecture and provides observations on recent trends. We also discuss the many innovations that have advanced in deep learning and discuss how these tools facilitate U-net. In addition, we review the different image modalities and application areas that have been enhanced by U-net.</div><div>In recent years, deep learning for health care is rapidly infiltrating and transforming medical fields thanks to the advances in computing power, data availability, and algorithm development. In particular, U-Net, a deep learning technique, has achieved remarkable success in medical image segmentation and has become one of the premier tools in this area. While the accomplishments of U-Net and other deep learning algorithms are evident, there still exist many challenges in medical image processing to achieve human-like performance. In this thesis, we propose a U-net architecture that integrates a residual skip connections and recurrent feedback with EfficientNet as a pretrained encoder. Residual connections help feature propagation in deep neural networks and significantly improve performance against networks with a similar number of parameters while recurrent connections ameliorate gradient learning. We also propose a second model that utilizes densely connected layers aiding deeper neural networks. And the proposed third model that incorporates fractal expansions to bypass diminishing gradients. EfficientNet is a family of powerful pretrained encoders that streamline neural network design. The use of EfficientNet as an encoder provides the network with robust feature extraction that can be used by the U-Net decoder to create highly accurate segmentation maps. The proposed networks are evaluated against state-of-the-art deep learning based segmentation techniques to demonstrate their superior performance.</div>

This diploma thesis deals with the application of deep neural networks with focus on image segmentation. The theoretical part contains a description of deep neural networks and a summary of widely used convolutional architectures for segmentation of objects from the image. Practical part of the work was devoted to testing of an existing network architectures. For this purpose, an open-source software library Tensorflow, implemented in Python programming language, was used. A frequent problem incorporating the use of convolutional neural networks is the requirement on large amount of input data. In order to overcome this obstacle a new data set, consisting of a combination of five freely available databases was created. The selected U-net network architecture was tested by first modification of the newly created data set. Based on the test results, the chosen network architecture has been modified. By these means a new network has been created achieving better performance in comparison to the original network. The modified architecture is then trained on a newly created data set, that contains images of different types taken with various fundus cameras. As a result, the trained network is more robust and allows segmentation of retina blood vessels from images with different parameters. The modified architecture was tested on the STARE, CHASE, and HRF databases. Results were compared with published segmentation methods from literature, which are based on convolutional neural networks, as well as classical segmentation methods. The created network shows a high success rate of retina blood vessels segmentation comparable to state-of-the-art methods.

This paper deals with classifying objects using deep neural networks. Whole scene segmentation was used as main algorithm for the classification purpose which works with video sequences and obtains information between two video frames. Optical flow was used for getting information from the video frames, based on which features maps of a~neural network are warped. Two neural network architectures were adjusted to work with videos and experimented with. Results of the experiments show, that using videos for image segmentation improves accuracy (IoU) compared to the same architecture working with images.

This thesis deals with possibilities of automatic segmentation of biomedical images. For the 3D image segmentation, a deep learning method has been proposed. In the work problems of network design, memory optimization method and subsequent composition of the resulting image are solved. The uniqueness of the method lies in 3D image processing on a GPU in combination with augmentation of training data and preservation of the output size with the original image. This is achieved by dividing the image into smaller parts with the overlay and then folding to the original size. The functionality of the method is verified on the segmentation of human brain tissue on magnetic resonance imaging, where it overcomes human accuracy when compared a specialist vs. specialist, and cell segmentation on a slices of the Drosophila brain from an electron microscope, where published results from the impacted paper are overcome.