Imagens codificadas por matrizes de intensidade são tipicamente representadas por grande quantidade de dados. Embora existam inúmeras abordagens para análise de imagens, o conhecimento sobre problemas específicos é raramente considerado. Este trabalho trata sobre problemas de análises de imagens cujas soluções dependem do conhecimento sobre os dados envolvidos na aplicação específica. Para isso, utiliza técnicas de mineração de dados para modelar as respostas humanas obtidas de experimentos psicofísicos. Dois problemas de análise de imagens são apresentados: (1) a análise de formas e (2) a análise pictórica. No primeiro problema (1), formas de neurônios da retina (neurônios ganglionares de gato) são segmentadas e seus contornos submetidos a uma calibração dos parâmetros de curvatura considerando a segmentação manual de um especialista. Outros descritores, tais como esqueletos multi-escalas são explorados para eventual uso e avaliação da abordagem. No segundo problema (2), a análise pictórica de imagens de home-pages serve para avaliar critérios estéticos a partir de medidas de complexidade, contraste e textura. O sistema generaliza as respostas por um experimento psicofísico realizados com humanos. Os resultados objetivos com as duas abordagens revelaram-se promissores, surpreendentes e com ampla aplicabilidade. / Images coded by intensity matrices typically involve large amount of data. Although image analysis approaches are diverse, knowledge about specific problems is rarely considered. This work is about image analysis problems whose solutions depend on the knowledge about the involved data. In order to do so data mining techniques are applied to model human response to psychophysical experiments. Two image analysis problems are addressed: (1) shape analysis; and (2) pictorial analysis. In the former, neuronal images (ganglion retinal cells of cat) are segmented and curvature parameters are calibrated to identify extremities and branches on the shape considering human segmentation as a reference. Descriptors such as multiscale skeletons are also explored for potential application or evaluations. In the second problem, a pictorial analysis of home-pages images feed an artificial aesthetics criteria evaluator based on complexity, contrast and texture features. The system models and generalizes the obtained human responses to psychophysical experiment. The results for these two approaches are promising, surprising and widely applicable.

No description available.

Imagens codificadas por matrizes de intensidade são tipicamente representadas por grande quantidade de dados. Embora existam inúmeras abordagens para análise de imagens, o conhecimento sobre problemas específicos é raramente considerado. Este trabalho trata sobre problemas de análises de imagens cujas soluções dependem do conhecimento sobre os dados envolvidos na aplicação específica. Para isso, utiliza técnicas de mineração de dados para modelar as respostas humanas obtidas de experimentos psicofísicos. Dois problemas de análise de imagens são apresentados: (1) a análise de formas e (2) a análise pictórica. No primeiro problema (1), formas de neurônios da retina (neurônios ganglionares de gato) são segmentadas e seus contornos submetidos a uma calibração dos parâmetros de curvatura considerando a segmentação manual de um especialista. Outros descritores, tais como esqueletos multi-escalas são explorados para eventual uso e avaliação da abordagem. No segundo problema (2), a análise pictórica de imagens de home-pages serve para avaliar critérios estéticos a partir de medidas de complexidade, contraste e textura. O sistema generaliza as respostas por um experimento psicofísico realizados com humanos. Os resultados objetivos com as duas abordagens revelaram-se promissores, surpreendentes e com ampla aplicabilidade. / Images coded by intensity matrices typically involve large amount of data. Although image analysis approaches are diverse, knowledge about specific problems is rarely considered. This work is about image analysis problems whose solutions depend on the knowledge about the involved data. In order to do so data mining techniques are applied to model human response to psychophysical experiments. Two image analysis problems are addressed: (1) shape analysis; and (2) pictorial analysis. In the former, neuronal images (ganglion retinal cells of cat) are segmented and curvature parameters are calibrated to identify extremities and branches on the shape considering human segmentation as a reference. Descriptors such as multiscale skeletons are also explored for potential application or evaluations. In the second problem, a pictorial analysis of home-pages images feed an artificial aesthetics criteria evaluator based on complexity, contrast and texture features. The system models and generalizes the obtained human responses to psychophysical experiment. The results for these two approaches are promising, surprising and widely applicable.

Biologically inspired logarithmic CMOS sensors offer high dynamic range imaging capabilities without the difficulties faced by linear imagers. By compressing dynamic range while encoding contrast information, they mimic the human visual system’s response to photo stimuli in fewer bits than those used in linear sensors. Despite this prospect, logarithmic sensors suffer poor image quality due to illumination dependent fixed pattern noise (FPN), making individual pixels appear up to 100 times brighter or darker. This thesis is primarily concerned with alleviating FPN in logarithmic imagers in a simple and convenient way while undertaking a system approach to its origin, distribution and effect on the quality of monochrome and colour images, after FPN correction. Using the properties of the Human visual system, I propose to characterise the errors arising from FPN in a perceptually significant manner by proposing an error measure, never used before. Logarithmic operation over a wide dynamic range is first characterised using a new model; yi j =aj +bj ln(exp sqrt(cj +djxi)−1), where yi j is the response of the sensor to a light stimulus xi and aj, bj, cj and dj are pixel dependent parameters. Using a proposed correction procedure, pixel data from a monochromatic sensor array is FPN corrected to approximately 4% error over 5 decades of illumination even after digitisation - accuracy equivalent to four times the human eyes ability to just notice an illumination difference against a uniform background. By evaluating how error affects colour, the possibility of indiscernible residual colour error after FPN correction, is analytically explored using a standard set of munsell colours. After simulating the simple FPN correction procedure, colour quality is analysed using a Delta E76 perceptual metric, to check for perceptual discrepancies in image colour. It is shown that, after quantisation, the FPN correction process yields 1−2 Delta E76 error units over approximately 5 decades of illumination; colour quality being imperceptibly uniform in this range. Finally, tone-mapping techniques, required to compress high dynamic range images onto the low range of standard screens, have a predominantly logarithmic operation during brightness compression. A new Logr'Gb' colour representation is presented in this thesis, significantly reducing computational complexity, while encoding contrast information. Using a well-known tone mapping technique, images represented in this new format are shown to maintain colour accuracy when the green colour channel is compressed to the standard display range, instead of the traditional luminance channel. The trade off between colour accuracy and computation in this tone mapping approach is also demonstrated, offering a low cost alternative for applications with low display specifications.

Video is a sequence of 2D images of the 3D world generated by a camera. As the camera moves relative to the real scene and elements of that scene themselves move, correlated frame-to-frame changes in the video images are induced. Humans easily identify such changes as scene motion and can readily assess attempts to quantify it. For a machine, the identification of the 2D frame-to-frame motion is difficult. This problem is addressed by the computer vision process of tracking. Tracking underpins the solution to the problem of augmenting general video sequences with artificial imagery, a staple task in the visual effects industry. The problem is difficult because tracking in general video sequences is complicated by the presence of non-rigid motion, repeated texture and arbitrary occlusions. Existing methods provide solutions that rely on imposing limitations on the scenes that can be processed or that rely on human artistry and hard work. I introduce new paradigms, frameworks and algorithms for overcoming the challenges of processing general video and thus provide solutions that fill the gap between the `automated' and `manual' approaches. The work is easily sectioned into three parts, which can be considered separately or taken together for dealing with video without limitations. The initial focus is on directly addressing practical issues of human interaction in the tracking process: a new solution is developed by explicitly incorporating the user into an interactive algorithm. It is a novel tracking system based on fast full-frame patch searching and high-speed optimal track determination. This approach makes only minimal assumptions about motion and appearance, making it suitable for the widest variety of input video. I detail an implementation of the new system using k-d trees and dynamic programming. The second distinct contribution is an important extension to tracking algorithms in general. It can be noted that existing tracking algorithms occupy a spectrum in their use of global motion information. Local methods are easily confused by occlusions, repeated texture and image noise. Global motion models offer strong predictions to see through these difficulties and have been used in restricted circumstances, but are defeated by scenes containing independently moving objects or modest levels of non-rigid motion. I present a well principled way of combining local and global models to improve tracking, especially in these highly problematic cases. By viewing rank-constrained tracking as a probabilistic model of 2D tracks instead of 3D motion, I show how one can obtain a robust motion prior that can be easily incorporated in any existing tracking algorithm. The development of the global motion prior is based on rank-constrained factorization of measurement matrices. A common difficulty comes from the frequent occurrence of occlusions in video, which means that the relevant matrices are often not complete due to missing data. This defeats standard factorization algorithms. To fully explain and understand the algorithmic complexities of factorization in this practical context, I present a common notation for the direct comparison of existing algorithms and propose a new family of hybrid approaches that combine the superb initial performance of alternation methods with the convergence power of the Newton algorithm. Together, these investigations provide a wide-ranging, yet coherent exploration of tracking non-rigid objects in video.

In modern society, cancer has become one of the most terrifying diseases because of its high and increasing death rate. The disease's deep impact demands extensive research to detect and eradicate it in all its forms. Breast cancer is one of the most common forms of cancer, and approximately one in nine women in the Western world will develop it over the course of their lives. Screening programmes have been shown to reduce the mortality rate, but they introduce an enormous amount of information that must be processed by radiologists on a daily basis. Computer Aided Diagnosis (CAD) systems aim to assist clinicians in their decision-making process, by acting as a second opinion and helping improve the detection and classification ratios by spotting very difficult and subtle cases. Although the field of cancer detection is rapidly developing and crosses over imaging modalities, X-ray mammography remains the principal tool to detect the first signs of breast cancer in population screening. The advantages and disadvantages of other imaging modalities for breast cancer detection are discussed along with the improvements and difficulties encountered in screening programmes. Remarkable achievements to date in breast CAD are equally presented. This thesis introduces original results for the detection of features from mammographic image analysis to improve the effectiveness of early cancer screening programmes. The detection of early signs of breast cancer is vital in managing such a fast developing disease with poor survival rates. Some of the earliest signs of cancer in the breast are the clusters of microcalcifications. The proposed method is based on image filtering comprising partial differential equations (PDE) for image enhancement. Subsequently, microcalcifications are segmented using characteristics of the human visual system, based on the superior qualities of the human eye to depict localised changes of intensity and appearance in an image. Parameters are set according to the image characteristics, which makes the method fully automated. The detection of breast masses in temporal mammographic pairs is also investigated as part of the development of a complete breast cancer detection tool. The design of this latter algorithm is based on the detection sequence used by radiologists in clinical routine. To support the classification of masses into benign or malignant, novel tumour features are introduced. Image normalisation is another key concept discussed in this thesis along with its benefits for cancer detection.

In this thesis, we investigate the problem of human layout estimation in unconstrained still images. This involves predicting the spatial configuration of body parts. We start our investigation with pictorial structure models and propose an efficient method of model fitting using skin regions. To detect the skin, we learn a colour model locally from the image by detecting the facial region. The resulting skin detections are also used for hand localisation. Our next contribution is a comprehensive dataset of 2D hand images. We collected this dataset from publicly available image sources, and annotated images with hand bounding boxes. The bounding boxes are not axis aligned, but are rather oriented with respect to the wrist. Our dataset is quite exhaustive as it includes images of different hand shapes and layout configurations. Using our dataset, we train a hand detector that is robust to background clutter and lighting variations. Our hand detector is implemented as a two-stage system. The first stage involves proposing hand hypotheses using complementary image features, which are then evaluated by the second stage classifier. This improves both precision and recall and results in a state-of-the-art hand detection method. In addition we develop a new method of non-maximum suppression based on super-pixels. We also contribute an efficient training algorithm for structured output ranking. In our algorithm, we reduce the time complexity of an expensive training component from quadratic to linear. This algorithm has a broad applicability and we use it for solving human layout estimation and taxonomic multiclass classification problems. For human layout, we use different body part detectors to propose part candidates. These candidates are then combined and scored using our ranking algorithm. By applying this bottom-up approach, we achieve accurate human layout estimation despite variations in viewpoint and layout configuration. In the multiclass classification problem, we define the misclassification error using a class taxonomy. The problem then reduces to a structured output ranking problem and we use our ranking method to optimise it. This allows inclusion of semantic knowledge about the classes and results in a more meaningful classification system. Lastly, we substantiate our ranking algorithm with theoretical proofs and derive the generalisation bounds for it. These bounds prove that the training error reduces to the lowest possible error asymptotically.

This thesis describes the development of methods for automatically obtaining coarse gaze direction estimates for pedestrians in surveillance video. Gaze direction estimates are beneficial in the context of surveillance as an indicator of an individual's intentions and their interest in their surroundings and other people. The overall task is broken down into two problems. The first is that of tracking large numbers of pedestrians in low resolution video, which is required to identify the head regions within video frames. The second problem is to process the extracted head regions and estimate the direction in which the person is facing as a coarse estimate of their gaze direction. The first approach for head tracking combines image measurements from HOG head detections and KLT corner tracking using a Kalman filter, and can track the heads of many pedestrians simultaneously to output head regions with pixel-level accuracy. The second approach uses Markov-Chain Monte-Carlo Data Association (MCMCDA) within a temporal sliding window to provide similarly accurate head regions, but with improved speed and robustness. The improved system accurately tracks the heads of twenty pedestrians in 1920x1080 video in real-time and can track through total occlusions for short time periods. The approaches for gaze direction estimation all make use of randomised decision tree classifiers. The first develops classifiers for low resolution head images that are invariant to hair and skin colours using branch decisions based on abstract labels rather than direct image measurements. The second approach addresses higher resolution images using HOG descriptors and novel Colour Triplet Comparison (CTC) based branches. The final approach infers custom appearance models for individual scenes using weakly supervised learning over large datasets of approximately 500,000 images. A Conditional Random Field (CRF) models interactions between appearance information and walking directions to estimate gaze directions for head image sequences.

Non-rigid image registration is an important tool for analysing morphometric differences in subjects with Alzheimer's disease from structural magnetic resonance images of the brain. This thesis describes a novel probabilistic approach to non-rigid registration of medical images, and explores the benefits of its use in this area of neuroimaging. Many image registration approaches have been developed for neuroimaging. The vast majority suffer from two limitations: Firstly, the trade-off between image fidelity and regularisation requires selection. Secondly, only a point-estimate of the mapping between images is inferred, overlooking the presence of uncertainty in the estimation. This thesis introduces a novel probabilistic non-rigid registration model and inference scheme. This framework allows the inference of the parameters that control the level of regularisation, and data fidelity in a data-driven fashion. To allow greater flexibility, this model is extended to allow the level of data fidelity to vary across space. A benefit of this approach, is that the registration can adapt to anatomical variability and other image acquisition differences. A further advantage of the proposed registration framework is that it provides an estimate of the distribution of probable transformations. Additional novel contributions of this thesis include two proposals for exploiting the estimated registration uncertainty. The first of these estimates a local image smoothing filter, which is based on the registration uncertainty. The second approach incorporates the distribution of transformations into an ensemble learning scheme for statistical prediction. These techniques are integrated into standard frameworks for morphometric analysis, and are demonstrated to improve the ability to distinguish subjects with Alzheimer's disease from healthy controls.

Poor drug penetration through tumor tissue has emerged as a fundamental obstacle to cancer therapy. The solid tumor microenvironment presents several physiological abnormalities which reduce the uptake of intravenously administered therapeutics, including leaky, irregularly spaced blood vessels, and a pressure gradient which resists transport of therapeutics from the bloodstream into the tumor. Because of these factors, a systemically administered anti-cancer agent is unlikely to reach 100% of cancer cells at therapeutic dosages, which is the efficacy required for curative treatment. The goal of this project is to use high-intensity focused ultrasound (HIFU) to enhance drug delivery via phenomena associated with acoustic cavitation. ‘Cavitation’ is the formation, oscillation, and collapse of bubbles in a sound field, and can be broadly divided into two types: ‘inertial’ and ‘stable’. Inertial cavitation involves violent bubble collapse and is associated with phenomena such as heating, fluid jetting, and broadband noise emission. Stable cavitation occurs at lower pressure amplitudes, and can generate liquid microstreaming in the bubble vicinity. It is the combination of fluid jetting and microstreaming which it is attempted to explore, control, and apply to the drug delivery problem in solid tumors. First, the potential for cavitation to enhance the convective transport of a model therapeutic into obstructed vasculature in a cell-free in vitro tumor model is evaluated. Transport is quantified using post-treatment image analysis of the distribution of a dye-labeled macromolecule, while cavitation activity is quantified by analyzing passively recorded acoustic emissions. The introduction of exogenous cavitation nuclei into the acoustic field is found to dramatically enhance both cavitation activity and convective transport. The strong correlation between inertial cavitation activity and drug delivery in this study suggested both a mechanism of action and the clinical potential for non-invasive treatment monitoring. Next, a flexible and efficient method to simulate numerically the microstreaming fields instigated by cavitating microbubbles is developed. The technique is applied to the problem of quantifying convective transport of a scalar quantity in the vicinity of acoustically cavitating microbubbles of various initial radii subject to a range of sonication parameters, yielding insight regarding treatment parameter choice. Finally, in vitro and in vivo models are used to explore the effect of HIFU on delivery and expression of a biologically active adenovirus. The role of cavitation in improving the distribution of adenovirus in porous media is established, as well as the critical role of certain sonication parameters in sustaining cavitation activity in vivo. It is shown that following intratumoral or intravenous co-injection of ultrasound contrast agents and adenovirus, both the distribution and expression of viral transgenes are enhanced in the presence of inertial cavitation. This ultrasound-based drug delivery system has the potential to be applied in conjunction with a broad range of macromolecular therapeutics to augment their bioavailability for cancer treatment. In order to reach this objective, further developmental work is recommended, directed towards improving therapeutic transducer design, using transducer arrays for treatment monitoring and mapping, and continuing the development of functionalized monodisperse cavitation nuclei.