The application of Image Processing to Autonomous driving has drawn significant attention in recently. However, the demanding nature of the image processing algorithms conveys a considerable burden to any conventional realtime implementation. On the other hand, the emergence of FPGAs has brought numerous facilities toward fast prototyping and implementation of ASICs so that an image processing algorithm can be designed, tested and synthesized in a relatively short period in comparison to traditional approaches. This thesis investigates the best combination of current algorithms to reach an optimum solution to the problem of lane detection and tracking, while aiming to fit the design to a minimal system. The proposed structure realizes three algorithms, namely Edge Detector, Hough Transform, and Kalman filter. For each module, the theoretical background is investigated and a detailed description of the realization is given followed by an analysis of both achievements and shortages of the design. It is concluded by describing the advantages of implementing this architecture and the use of these kinds of systems. / Tillämpningen av bildbehandling inom autonoma fordon har fått stor uppmärksamhet den senaste tiden. Emellertid förmedlar den krävande karaktären hos bildbehandlingsalgoritmerna en stor belastning på vilken konventionell realtidsimplementering som helst. Å andra sidan har framväxten av FPGAer medfört många möjligheter till snabb prototypering och implementering av ASICar så att en bildbehandlingsalgoritm kan utformas, testas och syntetiseras på relativt kort tid jämfört med traditionella tillvägagångssätt. Denna avhandling undersöker den bästa kombinationen av nuvarande algoritmer för att uppnå en optimal lösning på problemet med spårning och fildetektering, med målet att krympa designen till ett minimalt system. Den föreslagna strukturen realiserar tre algoritmer, nämligen Edge Detector, Hough Transform och Kalman filter. För varje modul undersöks den teoretiska bakgrunden och en detaljerad beskrivning av realiseringen ges följd av en analys av både fördelar och brister i konstruktionen. Avhandlingen avslutas med en beskrivning av fördelarna med att implementera lösningen på det sätt den görs och hur dessa system kan användas.

Com a crescente necessidade de segurança, o processo de identificação pessoal é cada vez mais exigido. A extração de características faciais é um passo importante quando se lida com interpretação visual automatizada no reconhecimento de faces humanas. Dentre as características faciais, os olhos são partes importantes no processo de reconhecimento, pois determinam o início da busca por outras características relevantes. Neste trabalho é apresentado um método de localização de olhos em imagens frontais de faces humanas. Este método é subdividido em duas partes. Primeiro, são identificados os possíveis candidatos a olhos usando a Transformada de Hough para círculos; depois é aplicada a Distância Euclidiana confirmando-se a localização pro biometria facial. / Personal identification process is an exigency for security systems. Facial feature extraction is a crucial step for automated visual interpretation in human face recognition. Withim all the facial features, the eyes are significantly parts for the recognition process, therefore they set up the start for another relevant feature search. In this work, we present a method for eyes locating in digital images of frontal human faces. This method is subdivided into two parts. First, we identify the possible eyes\'s candidates by Hough Transfor for circules, them we apply the Euclidian distance and calculate the eyes\'s position by facial biometric measurement.

Detecção automática de regiões planares em nuvens de pontos é um importante passo para muitas aplicações gráficas, de processamento de imagens e de visão computacional. Enquanto a disponibilidade de digitalizadores a laser e a fotografia digital tem nos permitido capturar nuvens de pontos cada vez maiores, técnicas anteriores para detecção de planos são computacionalmente caras, sendo incapazes de alcançar desempenho em tempo real para conjunto de dados contendo dezenas de milhares de pontos, mesmo quando a detecção é feita de um modo não determinístico. Apresentamos uma abordagem determinística para detecção de planos em nuvens de pontos não estruturadas que apresenta complexidade computacional O(n log n) no número de amostras de entrada. Ela é baseada em um método eficiente de votação para a transformada de Hough. Nossa estratégia agrupa conjuntos de pontos aproximadamente coplanares e deposita votos para estes conjuntos em um acumulador esférico, utilizando núcleos Gaussianos trivariados. Uma comparação com as técnicas concorrentes mostra que nossa abordagem é consideravelmente mais rápida e escala significativamente melhor que as técnicas anteriores, sendo a primeira solução prática para detecção determinística de planos em nuvens de pontos grandes e não estruturadas. / Automatic detection of planar regions in point clouds is an important step for many graphics, image processing, and computer vision applications. While laser scanners and digital photography have allowed us to capture increasingly larger datasets, previous techniques are computationally expensive, being unable to achieve real-time performance for datasets containing tens of thousands of points, even when detection is performed in a non-deterministic way. We present a deterministic technique for plane detection in unorganized point clouds whose cost is O(n log n) in the number of input samples. It is based on an efficient Hough-transform voting scheme and works by clustering approximately co-planar points and by casting votes for these clusters on a spherical accumulator using a trivariate Gaussian kernel. A comparison with competing techniques shows that our approach is considerably faster and scales significantly better than previous ones, being the first practical solution for deterministic plane detection in large unorganized point clouds.

A segmentação é uma etapa intermediária no registro e reconstrução 3D do pulmão. Geralmente, os métodos de segmentação são interativos e utilizam diferentes estratégias para combinar a expertise dos humanos e a velocidade e precisão dos computadores. A segmentação de imagens RM do pulmão é particularmente difícil devido à grande variação na qualidade da imagem. Dois métodos para a segmentação do contorno do pulmão são apresentados. No primeiro, uma análise individual de cada imagem da série de imagens RM é realizada, e a segmentação ocorre através de técnicas de limiarização e labeling. No segundo método, a respiração é associada a uma função respiração padrão, e através de técnicas de processamento de imagem 2D, detecção de bordas e transformada de Hough, padrões respiratórios são obtidos e, conseqüentemente, a posição dos pontos no tempo são estimados. Seqüências temporais de imagens RM são segmentadas, considerando a coerência no tempo. Desta forma, a silhueta do pulmão pode ser determinada em cada quadro, mesmo em quadros com bordas obscuras. A região do pulmão é segmentada em três etapas, neste método: uma máscara contendo a região do pulmão é criada a partir do resultado do primeiro método de segmentação; a transformada de Hough é aplicada exclusivamente aos pixels da máscara em diversos planos; o contorno do pulmão é extraído do resultado da transformada de Hough utilizando os contornos ativos. O formato da máscara pode ter uma grande variação, e a transformada de Hough modificada pode lidar com essa variação. Os resultados obtidos pelos dois métodos são comparados. / The segmentation of the lung is an intermediary step towards its registry and 3D reconstruction. Usually, segmentation methods are interactive and make use of different strategies to combine the expertise of the human and the computers accuracy and speed. Segmentation of lung magnetic resonance (MR) images is particularly difficult because of the large variation in image quality. Two methods for the lung contour segmentation are presented. In this first method, an individual analysis of each image in the series approach is taken, and the segmentation is made through thresholding and labeling techniques. In the second method, the breathing is associated to a standard respiratory function, and through 2D image processing, edge detection and Hough transform, respiratory patterns are obtained and, consequently, the position of points in time are estimated. Temporal sequences of MR images are segmented by considering the coherence in time. This way, the lung silhouette can be determined in every frame, even on frames with obscure edges. The lung region is segmented in three steps: a mask containing the lung region is created from the results of the first method; the Hough transform is applied exclusively to mask pixels in different planes; and the lung contour is created from the results of the Hough transform through active contours. The shape of the mask can have a large variation, and the modified Hough transform can handle such a shape variation. Results from both methods are compared.

Medical imaging has advanced at a tremendous rate since x-rays were discovered in 1895. Today, x-ray machines produce extremely high-quality images for radiologists to interpret. However, the methods of interpretation have only recently begun to be augmented by advances in computer technology. Computer aided diagnosis (CAD) systems that guide healthcare professionals to making the correct diagnosis are slowly becoming more prevalent throughout the medical field. Bone fractures are a relatively common occurrence. In most developed countries the number of fractures associated with age-related bone loss is increasing rapidly. Regardless of the treating physician's level of experience, accurate detection and evaluation of musculoskeletal trauma is often problematic. Each year, the presence of many fractures is missed during x-ray diagnosis. For a trauma patient, a mis-diagnosis can lead to ineffective patient management, increased dissatisfaction, and expensive litigation. As a result, detection of long-bone fractures is an important orthopaedic and radiologic problem, and it is proposed that a novel CAD system could help lower the miss rate. This thesis examines the development of such a system, for the detection of long-bone fractures. A number of image processing software algorithms useful for automating the fracture detection process have been created. The first algorithm is a non-linear scale-space smoothing technique that allows edge information to be extracted from the x-ray image. The degree of smoothing is controlled by the scale parameter, and allows the amount of image detail that should be retained to be adjusted for each stage of the analysis. The result is demonstrated to be superior to the Canny edge detection algorithm. The second utilises the edge information to determine a set of parameters that approximate the shaft of the long-bone. This is achieved using a modified Hough Transform, and specially designed peak and line endpoint detectors. The third stage uses the shaft approximation data to locate the bone centre-lines and then perform diaphysis segmentation to separate the diaphysis from the epiphyses. Two segmentation algorithms are presented and one is shown to not only produce better results, but also be suitable for application to all long-bone images. The final stage applies a gradient based fracture detection algorithm to the segmented regions. This algorithm utilises a tool called the gradient composite measure to identify abnormal regions, including fractures, within the image. These regions are then identified and highlighted if they are deemed to be part of a fracture. A database of fracture images from trauma patients was collected from the emergency department at the Flinders Medical Centre. From this complete set of images, a development set and test set were created. Experiments on the test set show that diaphysis segmentation and fracture detection are both performed with an accuracy of 83%. Therefore these tools can consistently identify the boundaries between the bone segments, and then accurately highlight midshaft long-bone fractures within the marked diaphysis. Two of the algorithms---the non-linear smoothing and Hough Transform---are relatively slow to compute. Methods of decreasing the diagnosis time were investigated, and a set of parallelised algorithms were designed. These algorithms significantly reduced the total calculation time, making use of the algorithm much more feasible. The thesis concludes with an outline of future research and proposed techniques that---along with the methods and results presented---will improve CAD systems for fracture detection, resulting in more accurate diagnosis of fractures, and a reduction of the fracture miss rate.

<p>Feature extraction from a tensor based local image representation introduced by Knutsson in [37] is discussed. The tensor representation keeps statements of structure, certainty of statement and energy separate. Further processing for obtaining new features also having these three entities separate is achieved by the use of a new concept, tensor field filtering. Tensor filters for smoothing and for extraction of circular symmetries are presented and discussed in particular. These methods are used for corner detection and extraction of more global features such as lines in images. A novel method for grouping local orientation estimates into global line parameters is introduced. The method is based on a new parameter space, the Möbius Strip parameter space, which has similarities to the Hough transform. A local centroid clustering algorithm is used for classification in this space. The procedure automatically divides curves into line segments with appropriate lengths depending on the curvature. A linked list structure is built up for storing data in an efficient way.</p> / Ogiltigt nummer / annan version: I publ. nr 290:s ISBN: 91-7870-815-X.

Accurate 3D models of natural environments are important for many modelling and simulation applications, for both civilian and military purposes. When building 3D models from high resolution data acquired by an airborne laser scanner it is de-sirable to separate and classify the data to be able to process it further. For example, to build a polygon model of a building the samples belonging to the building must be found. In this thesis we have developed, implemented (in IDL and ENVI), and evaluated algorithms for classification of buildings, vegetation, power lines, posts, and roads. The data is gridded and interpolated and a ground surface is estimated before the classification. For the building classification an object based approach was used unlike most classification algorithms which are pixel based. The building classifica-tion has been tested and compared with two existing classification algorithms. The developed algorithm classified 99.6 % of the building pixels correctly, while the two other algorithms classified 92.2 % respective 80.5 % of the pixels correctly. The algorithms developed for the other classes were tested with thefollowing result (correctly classified pixels): vegetation, 98.8 %; power lines, 98.2 %; posts, 42.3 %; roads, 96.2 %.

Advances in hardware and pattern recognition techniques, along with the widespread utilization of remote sensing satellites, have urged the development of automatic target detection systems. Automatic detection of airports is particularly essential, due to the strategic importance of these targets. In this thesis, a detection method is proposed for airport runways, which is the most distinguishing element of an airport. This method, which operates on large optical satellite images, is composed of a segmentation process based on textural properties, and a runway shape detection stage. In the segmentation process, several local textural features are extracted including not only low level features such as mean, standard deviation of image intensity and gradient, but also Zernike Moments, Circular-Mellin Features, Haralick Features, as well as features involving Gabor Filters, Wavelets and Fourier Power Spectrum Analysis. Since the subset of the mentioned features, which have a role in the discrimination of airport runways from other structures and landforms, cannot be predicted, Adaboost learning algorithm is employed for both classification and determining the feature subset, due to its feature selector nature. By means of the features chosen in this way, a coarse representation of possible runway locations is obtained, as a result of the segmentation operation. Subsequently, the runway shape detection stage, based on a novel form of Hough Transform, is performed over the possible runway locations, in order to obtain final runway positions. The proposed algorithm is examined with experimental work using a comprehensive data set consisting of large and high resolution satellite images and successful results are achieved.

<p>Accurate 3D models of natural environments are important for many modelling and simulation applications, for both civilian and military purposes. When building 3D models from high resolution data acquired by an airborne laser scanner it is de-sirable to separate and classify the data to be able to process it further. For example, to build a polygon model of a building the samples belonging to the building must be found.</p><p>In this thesis we have developed, implemented (in IDL and ENVI), and evaluated algorithms for classification of buildings, vegetation, power lines, posts, and roads. The data is gridded and interpolated and a ground surface is estimated before the classification. For the building classification an object based approach was used unlike most classification algorithms which are pixel based. The building classifica-tion has been tested and compared with two existing classification algorithms. </p><p>The developed algorithm classified 99.6 % of the building pixels correctly, while the two other algorithms classified 92.2 % respective 80.5 % of the pixels correctly. The algorithms developed for the other classes were tested with thefollowing result (correctly classified pixels): vegetation, 98.8 %; power lines, 98.2 %; posts, 42.3 %; roads, 96.2 %.</p>

Three-dimensional endothelial cell sprouting assay (3D-ECSA) exhibits differentiation of endothelial cells into sprouting structures inside a 3D matrix of collagen I. It is a screening tool to study endothelial cell behavior and identification of angiogenesis inhibitors. The shape and size of an EC spheroid (aggregation of ~ 750 cells) is important with respect to its growth performance in presence of angiogenic stimulators. Apparently, tubules formed on malformed spheroids lack homogeneity in terms of density and length. This requires segregation of well formed spheroids from malformed ones to obtain better performance metrics. We aim to develop and validate an automated imaging software analysis tool, as a part of a High-content High throughput screening (HC-HTS) assay platform, to exploit 3D-ECSA as a differential HTS assay. We present a solution using Circular Hough Transform to detect a nearly perfect spheroid as per its circular shape in a 2D image. This successfully enables us to differentiate and separate good spheroids from the malformed ones using automated test bench.