The objective of this thesis is to create Gaussian 1D filters with rotating kernel theory which enables to program algorithm for noise reduction and beam structure highlighting in a digital picture of the solar corona. A fragment of original picture of solar corona and of pictures filtred by this algorithm is in the enclosure.

Orientador: José Mario De Martino / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Elétrica e de Computação / Made available in DSpace on 2018-08-26T21:43:57Z (GMT). No. of bitstreams: 1 Costa_PaulaDornhoferParo_D.pdf: 15894797 bytes, checksum: 194a20ae502dfc7198a008d576e23e4c (MD5) Previous issue date: 2015 / Resumo: O desenvolvimento da tecnologia de animação facial busca atender uma demanda crescente por aplicações envolvendo assistentes, vendedores, tutores e apresentadores de notícias virtuais; personagens realistas de videogames, agentes sociais e ferramentas para experimentos científicos em psicologia e ciências comportamentais. Um aspecto relevante e desafiador no desenvolvimento de cabeças falantes, ou "talking heads", é a reprodução realista dos movimentos articulatórios da fala combinados aos elementos de comunicação não-verbal e de expressão de emoções. Este trabalho presenta uma metodologia de síntese de animação facial baseada em imagens, ou animação facial 2D, que permite a reprodução de uma ampla gama de estados emocionais de fala expressiva, além de suportar a modulação de movimentos da cabeça e o controle de elementos faciais tais como o piscar de olhos e o arqueamento de sobrancelhas. A síntese da animação utiliza uma base de imagens-protótipo que são processadas para obtenção dos quadros-chave da animação. Os pesos utilizados para a combinação das imagens-protótipo são derivados de um modelo estatístico de aparência e formas, construído a partir de um conjunto de imagens de treinamento extraídas de um corpus audiovisual de uma face real. A síntese das poses-chave é guiada pela transcrição fonética temporizada da fala a ser animada e pela informação do estado emocional almejado. As poses-chave representam visemas dependentes de contexto fonético que implicitamente modelam os efeitos da coarticulação na fala visual. A transição entre poses-chave adjacentes é realizada por um algoritmo de metamorfose não-linear entre imagens. As animações sintetizadas aplicando-se a metodologia proposta foram avaliadas por meio de avaliação perceptual de reconhecimento de emoções. Dentre as contribuições deste trabalho encontra-se a construção de uma base de dados de vídeo e captura de movimento para fala expressiva em português do Brasil / Abstract: The facial animation technology experiences an increasing demand for applications involving virtual assistants, sellers, tutors and newscasters; lifelike game characters, social agents, and tools for scientific experiments in psychology and behavioral sciences. A relevant and challenging aspect of the development of talking heads is the realistic reproduction of the speech articulatory movements combined with the elements of non-verbal communication and the expression of emotions. This work presents an image-based, or 2D, facial animation synthesis methodology that allows the reproduction of a wide range of expressive speech emotional states and also supports the modulation of head movements and the control of face elements, like the blinking of the eyes and the raising of the eyebrows. The synthesis of the animation uses a database of prototype images which are combined to produce animation keyframes. The weights used for combining the prototype images are derived from a statistical active appearance model (AAM), which is built from a set of sample images extracted from an audio-visual corpus of a real face. The generation of the animation keyframes is driven by the timed phonetic transcription of the speech to be animated and the desired emotional state. The keyposes consist of expressive context-dependent visemes that implicitly model the speech coarticulation effects. The transition between adjacent keyposes is performed through a non-linear image morphing algorithm. To evaluate the synthesized animations, a perceptual evaluation based on the recognition of emotions was performed. Among the contributions of the work is also the building of a database of expressive speech video and motion capture data for Brazilian Portuguese / Doutorado / Engenharia de Computação / Doutora em Engenharia Elétrica

Orientador: Carlos Alberto de Souza Filho / Tese (doutorado) - Universidade Estadual de Campinas, Instituto de Geociências / Made available in DSpace on 2018-08-27T12:53:23Z (GMT). No. of bitstreams: 1 Scafutto_RebeccadelPapaMoreira_D.pdf: 30942080 bytes, checksum: c174451bbb1ba49a30f8e98bee9743e4 (MD5) Previous issue date: 2015 / Resumo: No cenário econômico atual, as matrizes energéticas são peças fundamentais nas movimentações de mercado. O petróleo e produtos derivados, em especial, estão relacionados a diversas áreas da economia e, tendo em vista que a abundância destes produtos condiciona as variações na estabilidade econômica global, a descoberta de novos reservatórios é de extrema importância. Contudo, a descoberta e exploração de novos campos de petróleo estão vinculadasàa elevado risco de contaminação ambiental, sendo necessário também o monitoramento e remediação das áreas contaminadas. Apesar do grande avanço e pesquisas relacionadas às atividades petrolíferas em regiões marítimas (offshore), a identificação precoce de ocorrências de petróleo oriundas de exsudações naturais ou vazamentos acidentais em regiões continentais (onshore) é ainda um tema pouco explorado na literatura. A localização e mapeamento de áreas contaminadas, seja por micro exsudações ou vazamentos, podem direcionar as operações de exploração petrolífera. Por outro lado, as ocorrências superficiais de HCs podem indicar também a presença de danos em oleodutos e a localização de pequenos vazamentos pode guiar as operações contenção, reparo e limpeza. Neste cenário, as ferramentas de sensoriamento remoto (SR) hiperespectral são uma opção extremamente promissora para o desenvolvimento de técnicas que auxiliem simultaneamente na descoberta de novos reservatórios e na identificação precoce de vazamentos. A elevada resolução espectral dos dados hiperespectrais permitem a localização direta e monitoramento da dispersão de vazamentos, gerando informações valiosas, de baixo custo, adquiridas em um curto espaço de tempo que podem auxiliar nas operações de exploração ou contenção. Simultaneamente, em função da elevada quantidade de dados gerada por estes equipamentos, o processamento dos dados espectrais com técnicas de análise multivariada de dados apresenta potencial para alavancar programas de prospecção voltados a descoberta de novas ocorrências de petróleo, tornando-os mais rápidos e eficientes. Sendo assim, o objetivo deste trabalho foi o desenvolvimento de metodologias que auxiliem na detecção direta, mapeamento remoto e classificação de áreas continentais afetadas por exsudações naturais de petróleo ou vazamentos de hidrocarbonetos. Para atingir os objetivos propostos, as propriedades de reflectância espectral de misturas de substratos minerais diversos impregnados com etanol, diesel, gasolina e óleos crus de variadas densidades foram analisadas dentro do intervalo do Infravermelho Próximo e de Ondas Curtas; através de três experimentos controlados realizados em laboratório e campo. Os primeiros dois experimentos elaborados em laboratório permitiram a construção de uma ampla biblioteca espectral de referência, de substratos minerais impregnados com hidrocarbonetos, proposta pela primeira vez na literatura. O terceiro experimento consistiu em um estudo de caso no qual os dados gerados com as bibliotecas foram utilizados para a identificação e mapeamento de regiões continentais impregnadas com hidrocarbonetos, através da aplicação de algoritmos de classificação espectral. As técnicas de espectroscopia, análise multivariada e mapeamento espectral aplicadas aos dados hiperespectrais investigados nesta pesquisa demonstraram, num conjunto de experimentos originais, a possibilidade de caracterizar remotamente substratos minerais impregnados com HCs. Considerando-se os resultados positivos obtidos com experimentos controlados, as bibliotecas espectrais e modelos desenvolvidos nesta pesquisa têm potencial inovador para auxiliar na caracterização remota de HCs, bem como o uso de dados aerotransportados hiperespectrais para a prospecção petrolífera e monitoramento ambiental em regiões continentais. O desenvolvimento desta técnica, aliada a outras ferramentas exploratórias, abre caminhos para descoberta de novos campos de exploração onshore, bem como para o monitoramento ambiental de regiões afetadas por vazamentos de dutos em subsuperfície através de uma ferramenta rápida, eficaz e de baixo custo operacional / Abstract: In the current economic scenario, energy matrices are essential pieces in the market. Petroleum and derivatives, in particular, are related to several branches of the economy. Once the availability of these products determines variations in the stability of global economy, the discovery of new reservoirs is extremely important. However, the discovery and exploration of new oil fields are related to high risks of environmental contamination, making the monitoring and remediation of contaminated areas required. Despite the advance and increasing researches related to oil activities in offshore areas, the early identification of oil occurrences arising from natural seepage or accidental leaks in continental regions (onshore) is still a topic poorly studied. The location and mapping of contaminated areas, by either micro seepage or leaks, can lead oil exploration operations. On the other hand, occurrences of hydrocarbons (HCs) in the surface may also indicate the presence of cracks in pipelines and the identification of small leaks can guide operations of contention, restoration and cleaning. In this scenario, hyperspectral remote sensing tools (SR) appear as a promising option for the development of techniques to support both the discovery of new reservoirs and the early identification of leaks. The high spectral resolution of hyperspectral data allows the direct location and monitoring of oil occurrences, generating valuable and low cost information in real-time that can assist in exploration or containment operations. At the same time, due to the high amount of data generated by these equipment, the processing of spectral data with chemometric techniques has the potential to leverage exploration programs focused on the discovery of new oil occurrences, making them faster and more efficient. Therefore, the aim of this research was the development of new methodologies to assist in the direct detection, remote mapping and classification of continental areas affected by natural oil seepage or leaks. To achieve the proposed objectives, the spectral reflectance properties of mixtures of various mineral substrates impregnated with ethanol, diesel, gasoline and crude oils of varying densities were analyzed within the range of Near and Shortwave Infrared; through three controlled experiments performed in the laboratory and field. The first two experiments developed in laboratory enabled the building of a wide reference spectral library of mineral substrates impregnated with hydrocarbons, first proposed in the literature. The third experiment consisted of a case study in which the generated libraries were used for the remote identification and mapping of continental sites impregnated with HCs, through the application of spectral classification algorithms. The processing of hyperspectral data with techniques of spectroscopy, chemometry and spectral mapping investigated in this study, demonstrated the possibility to remotely characterize mineral substrates impregnated with HCs. Considering the optimistic results of controlled experiments, the spectral libraries and models developed in this research have the potential to assist in the remote characterization of hydrocarbons, as well as the use of airborne hyperspectral data to support oil exploration and environmental monitoring in continental regions. The development of this technique, combined with other exploratory tools, enables the discovery of new onshore exploration fields, as well as the environmental monitoring of onshore areas affected by pipeline leakages through a fast, effective and low operating cost too / Doutorado / Geologia e Recursos Naturais / Doutora em Ciências

The objective of this thesis is to create Gaussian 1D filters with rotating kernel theory which enables to program algorithm for noise reduction and beam structure highlighting in a digital picture of the solar corona.

This thesis deals with the design of a hardware acceleration unit for digital image filter design using coevolutionary algorithms. The first part introduces reconfigurable logic device technology that the acceleration unit is based on. The theoretical part also briefly characterizes evolutionary and coevolutionary algorithms, their principles and applications. Traditional image filter designs are compared with the biologically inspired design methods. The hardware unit presented in this thesis exploits dual MicroBlaze system extended by custom peripherals to accelerate cartesian genetic programming. The coevolutionary image filter design is accelerated up to 58 times. The hardware platform functionality in the task of impulse noise filter design and edge detector design has been empirically analyzed.

<p>Predložena je multiveličinska obrada kojom je moguće objediniti pojačanje<br />vidljivosti detalja kao i poboljšanje kontrasta radiografskih snimaka, kojim<br />se uspostavlja ravnoteža između kontrasta slabo i jasno vidljivih struktura.<br />Pored toga, obradom je obuhvaćeno i smanjenje globalnog kontrasta, čime je<br />moguće dodatno naglasiti lokalne strukture. Drugim rečima, predložena je<br />multiveličinska obrada koja integriše sve korake poboljšanja vidljivosti<br />anatomskih struktura. Predložena obrada razmotrena je u okviru razvijenog<br />algoritamskog okvira koji sadrži korake obrade radiografskih snimaka koji su<br />potrebni da bi se sirov signal, dobijen od strane detektora zračenja, obradio<br />i time pripremio za prikazivanje lekarima. Svaki od koraka obrade je<br />analiziran i predložena su originalna rešenja, kao i poboljšanja postojećih<br />pristupa. Evaluacijom je pokazano se integrisanom obradom postižu rezultati<br />koji prevazilaze one koji se dobijaju savremenom vrhunskom obradom, kao i da je<br />celokupni proces obrade moguće kontrolisati sa samo dva operativna<br />parametra. Da bi se upotpunila sveobuhvatna analiza procesa obrade<br />radiografskih snimaka, u disertaciji je razmotreno i uklanjanje artefakata<br />nastalih obradom, kao i mogućnost ubrzanja obrade radiografskih snimaka. Za<br />oba problema su ponuđena originalna rešenja čija je efikasnost<br />eksperimentalno potvrđena.</p> / <p>The thesis focuses on digital radiography image processing. Multi-scale processing is<br />proposed, which unifies detail visibility enhancement, local contrast enhancement and<br />global contrast reduction, thus enabling additional amplification of local structures. In<br />other words, the proposed multi-scale image processing integrates all steps of<br />anatomical structures visibility enhancement. For the purpose of the proposed<br />anatomical structures visibility enhancement analysis, a processing framework was<br />developed. The framework consists of several stages, used to process the image from<br />its raw form (signal obtained from the radiation detector), to the state where it will be<br />presented to the medical diagnostician. Each stage is analyzed and for each an<br />original solution or an improvement of an existing approach was proposed. Evaluation<br />has shown that integrated processing provides results which surpass state-of-the-art<br />processing methods, and that the entire processing pipeline can be controlled using<br />just two parameters. In order to complete the comprehensive analysis of radiography<br />image processing, processing artifacts removal and radiography image processing<br />acceleration are analyzed in the thesis. Both issues are addressed through original<br />solutions whose efficiency is experimentally confirmed.</p>

<p>To begin with, we describe a project in which three screens for Cyan, Magenta, and Yellow colorants were designed jointly using the Direct Binary Search algorithm (DBS). The screen set generated by the algorithm can be used to halftone color images easily and quickly. The halftoning results demonstrate that by utilizing the screen sets, it is possible to obtain high-quality color halftone images while significantly reducing computational complexity.</p> <p>Our next research focuses on defect detection and quality assessment of printed images. We measure and analyze macro-uniformity, banding, and color plane misregistration. For these three defects, we designed different pipelines for them and developed a series of digital image processing and computer vision algorithms for the purpose of quantifying and evaluating these printed image defects. Additionally, we conduct a human psychophysical experiment to collect perceptual assessments and use machine learning approaches to predict image quality scores based on human vision.</p> <p>We study modern deep convolutional neural networks for image denoising and propose a network designed for AWGN image denoising. </p> <p>Our network removes the bias at each layer to achieve the benefits of scaling invariant network; additionally, it implements a mix loss function to boost performance. We train and evaluate our denoising results using PSNR, SSIM, and LPIPS, and demonstrate that our results achieve impressive performance on both objective and subjective IQA assessments.</p>

Diese Arbeit befasst sich mit der automatisierten Analyse der Ausrichtungen von Neuronenfasern in Mikroskopiebildern des menschlichen Gehirns. Für eine solche Analyse wurden vom Paul-Flechsig-Institut für Hirnforschung Leipzig (PFI) Fluoreszenzbilddaten zur Verfügung gestellt. Um für diese Daten Faserausrichtungen zu ermitteln, werden drei Schritte durchgeführt: Neuronenfasern werden hervorgehoben, bzw. freigestellt, es werden Orientierungen zu diesen zugeordnet und die hierdurch ermittelten Ergebnisse werden visualisiert. Es werden für jeden dieser Schritte mehrere Verfahren der klassischen Bildverarbeitung vorgestellt und die Auswirkung verschiedener Parameter auf deren Ergebnisse untersucht. Betrachtet werden Verfahren zur Kontrasterhöhung, Gauß-Filter, auf Hessematrizen basierende Filter, Berechnung von Phasenübereinstimmung und eine Wavelet-Transformation. Alle während dieser Arbeit vorgenommenen Implementierungen stehen als Python-Skripte auf GitHub (https://github.com/saphyll/fiber-orientation) zur Verfügung.:Einleitung 1. Grundlagen 1.1 Datengrundlage 1.2 Architektur 1.3 Grundlagen der Bildverarbeitung 1.3.1 Histogramme 1.3.2 Konvolution 1.3.3 Gaußkernel 1.3.4 Hessematrix und Eigenvektoren 1.4.5 Fourier-Transformation 2. Faseranalyse in 2D 2.1 Hervorhebung von Fasern 2.1.1 Histogram Equalization 2.1.2 Gauß-Filter 2.1.3 Hessematrix-basierte Filter 2.1.4 Phase Congruency 2.1.5 Isotropic Undecimated Wavelet Transform 2.2 Analyse und Visualisierung von Faserrichtungen 2.2.1 Richtungshistogramme 2.2.2 Kacheln 2.2.3 Direkte Ergebnisbilder 3. Zusammenfassung und Ausblick / This thesis covers the automated analysis of fiber orientations in microscopic images of the human brain in regard to data provided by the Paul Flechsig Institute of Brain Research Leipzig (PFI). For the retrieval of information about fiber orientations, three steps are used: An enhancement of fiber visibility and definition, an assignment of orientations to those fibers and a visualisation of fibers and their orientations. Multiple methods from classical image processing are presented for each of these steps and are evaluated according to the available data. These methods include contrast enhancement, gaussian filters, hessian filters, calculation of phase congruency and a wavelet transformation. All implementations resulting from this thesis are available as Python scripts on GitHub (https://github.com/saphyll/fiber-orientation).:Einleitung 1. Grundlagen 1.1 Datengrundlage 1.2 Architektur 1.3 Grundlagen der Bildverarbeitung 1.3.1 Histogramme 1.3.2 Konvolution 1.3.3 Gaußkernel 1.3.4 Hessematrix und Eigenvektoren 1.4.5 Fourier-Transformation 2. Faseranalyse in 2D 2.1 Hervorhebung von Fasern 2.1.1 Histogram Equalization 2.1.2 Gauß-Filter 2.1.3 Hessematrix-basierte Filter 2.1.4 Phase Congruency 2.1.5 Isotropic Undecimated Wavelet Transform 2.2 Analyse und Visualisierung von Faserrichtungen 2.2.1 Richtungshistogramme 2.2.2 Kacheln 2.2.3 Direkte Ergebnisbilder 3. Zusammenfassung und Ausblick

Every year wildfires threaten or destroy ecological habitats, man-made infrastructure and people’s lives. Additionally millions of dollars are spent each year trying to prevent and control these fires. Ideally if a wildfire can be detected before it rages out of control it can be extinguished and avoid large scale devastation. Traditional manned fire lookout towers are neither cost effective nor particularly efficient at detecting wildfire. It is proposed that temporal filtering can be used to isolate the signals created at the beginnings of potential wildfires. Temporal filtering can remove any background image and any periodic signals created by the camera movement. Once typical signals are analyzed, digital filters can be designed to pass fire signals while blocking the unwanted signals. The temporal filter passes only fire signals and signals generated by moving objects. These objects can be distinguished from each other by analyzing the objects mid and long wave energy profile. This algorithm is tested on 17 data sources and its results analyzed.

Synthetic Aperture Radar (SAR) has come into widespread use in several civilian and military applications. The focus of this paper is the military application of imaging point targets captured by an airborne SAR platform. Using the traditional SAR method of determining target motion by analyzing the difference between subsequent images takes a relatively large amount of processing resources. Using methods in this thesis, target motion can be estimated before even a single image is obtained, reducing the amount of time and power used by a significantly large amount. This thesis builds on work done by Brain Zaharri and David So. Brain Zaharri successfully created a SAR simulation that accurately models the airborne SAR system capturing data of a target space using the Range Doppler Algorithm (RDA). David So extended this work by adding functionality to determine target velocity in the range and azimuth directions by processing the intermittent data created by the several steps of Brian Zaharri’s simulation. This thesis shows further extensions of processing the intermittent data using unique methods. The methods in this thesis successfully demonstrate the ability to quickly and accurately estimate target position, velocity, and acceleration without the need for using multiple SAR images. Target motion in the range direction is detected without using any part of the RDA, while the azimuth direction cuts out several steps, including the range compression phase and the range cell migration correction. Removing these unneeded processing steps dramatically decreases target motion data acquisition time. Both Brian Zaharri’s and David So’s work, along with this thesis, are part of the Cal Poly SAR Automatic Target Recognition (ATR) group of projects, which is sponsored by Raytheon Space & Airborne Systems Division. Because U.S. military SAR data remains classified, the Cal Poly SAR ATR projects addresses the need to educate researchers on the processing of SAR data.