In this research work, we develop an obstacle detection and emergency exit sign recognition system on a mobile phone by extending the feature from accelerated segment test detector with Harris corner filter. The first step often required for many vision based applications is the detection of objects of interest in an image. Hence, in this research work, we introduce emergency exit sign detection method using colour histogram. The hue and saturation component of an HSV colour model are processed into features to build a 2D colour histogram. We backproject a 2D colour histogram to detect emergency exit sign from a captured image as the first task required before performing emergency exit sign recognition. The result of classification shows that the 2D histogram is fast and can discriminate between objects and background with accuracy. One of the challenges confronting object recognition methods is the type of image feature to compute. In this work therefore, we present two feature detectors and descriptor methods based on the feature from accelerated segment test detector with Harris corner filter. The first method is called Upright FAST-Harris and binary detector (U-FaHB), while the second method Scale Interpolated FAST-Harris and Binary (SIFaHB). In both methods, feature points are extracted using the accelerated segment test detectors and Harris filter to return the strongest corner points as features. However, in the case of SIFaHB, the extraction of feature points is done across the image plane and along the scale-space. The modular design of these detectors allows for the integration of descriptors of any kind. Therefore, we combine these detectors with binary test descriptor like BRIEF to compute feature regions. These detectors and the combined descriptor are evaluated using different images observed under various geometric and photometric transformations and the performance is compared with other detectors and descriptors. The results obtained show that our proposed feature detector and descriptor method is fast and performs better compared with other methods like SIFT, SURF, ORB, BRISK, CenSurE. Based on the potential of U-FaHB detector and descriptor, we extended it for use in optical flow computation, which we termed the Nearest-flow method. This method has the potential of computing flow vectors for use in obstacle detection. Just like any other new methods, we evaluated the Nearest flow method using real and synthetic image sequences. We compare the performance of the Nearest-flow with other methods like the Lucas and Kanade, Farneback and SIFT-flow. The results obtained show that our Nearest-flow method is faster to compute and performs better on real scene images compared with the other methods. In the final part of this research, we demonstrate the application potential of our proposed methods by developing an obstacle detection and exit sign recognition system on a camera phone and the result obtained shows that the methods have the potential to solve this vision based object detection and recognition problem.

<p>Doktorska disertacija izučava primenu fazi operacija, prvenstveno agregacionih operatora na funkcije rastojanja i metrike. Originalan doprinos teze je u konstrukciji novih funkcija rastojanja i metrika primenom agregacionih operatora na neke polazne funkcije rastojanja i metrike. Za neke tipove agregacionih operatora i polaznih funkcija rastojanja i metrika su ispitane osobine ovako konstruisanih funkcija rastojanja i metrika. Za neke od njih su ispitane performanse pri primeni u segmentaciji slike &bdquo;Fuzzy c-means&ldquo; algoritmom.</p> / <p>This thesis studies application of fuzzy operations, especially aggregation operators, on distance functions and metrics. The contribution of the thesis is construction of new distance functions and metrics by application of aggregation operators on some basic distance functions and metrics. For some types of aggregation operators and basic distance functions and metrics, properties of distance functions and metrics constructed in this way are analyzed. For some of them, performances in application in Fuzzy c-means algorithm are analyzed.</p>

<p>Many different approaches have been taken towards solving the stereo correspondence problem and great progress has been made within the field during the last decade. This is mainly thanks to newly evolved global optimization techniques and better ways to compute pixel dissimilarity between views. The most successful algorithms are based on approaches that explicitly model smoothness assumptions made about the physical world, with image segmentation and plane fitting being two frequently used techniques.</p><p>Within the project, a survey of state of the art stereo algorithms was conducted and the theory behind them is explained. Techniques found interesting were implemented for experimental trials and an algorithm aiming to achieve state of the art performance was implemented and evaluated. For several cases, state of the art performance was reached.</p><p>To keep down the computational complexity, an algorithm relying on local winner-take-all optimization, image segmentation and plane fitting was compared against minimizing a global energy function formulated on pixel level. Experiments show that the local approach in several cases can match the global approach, but that problems sometimes arise – especially when large areas that lack texture are present. Such problematic areas are better handled by the explicit modeling of smoothness in global energy minimization.</p><p>Lastly, disparity estimation for image sequences was explored and some ideas on how to use temporal information were implemented and tried. The ideas mainly relied on motion detection to determine parts that are static in a sequence of frames. Stereo correspondence for sequences is a rather new research field, and there is still a lot of work to be made.</p>

Image segmentation is one of the most difficult tasks in image processing. Segmentation algorithms are generally based on searching a region where pixels share similar gray level intensity and satisfy a set of defined criteria. However, the segmented region cannot be used directly for partial image retrieval. In this dissertation, a Contour Based Image Structure (CBIS) model is introduced. In this model, images are divided into several objects defined by their bounding contours. The bounding contour structure allows individual object extraction, and partial object matching and retrieval from a standard CBIS image structure. The CBIS model allows the representation of 3D objects by their bounding contours which is suitable for parallel implementation particularly when extracting contour features and matching them for 3D images require heavy computations. This computational burden becomes worse for images with high resolution and large contour density. In this essence we designed two parallel algorithms; Contour Parallelization Algorithm (CPA) and Partial Retrieval Parallelization Algorithm (PRPA). Both algorithms have considerably improved the performance of CBIS for both contour shape matching as well as partial image retrieval. To improve the effectiveness of CBIS in segmenting images with inhomogeneous backgrounds we used the phase congruency invariant features of Fourier transform components to highlight boundaries of objects prior to extracting their contours. The contour matching process has also been improved by constructing a fuzzy contour matching system that allows unbiased matching decisions. Further improvements have been achieved through the use of a contour tailored Fourier descriptor to make translation and rotation invariance. It is proved to be suitable for general contour shape matching where translation, rotation, and scaling invariance are required. For those images which are hard to be classified by object contours such as bacterial images, we define a multi-level cosine transform to extract their texture features for image classification. The low frequency Discrete Cosine Transform coefficients and Zenike moments derived from images are trained by Support Vector Machine (SVM) to generate multiple classifiers.

The automatic extraction of knowledge from the world by a robotic system as human beings interpret their environment through their senses is still an unsolved task in Artificial Intelligence. A robotic agent is in contact with the world through its sensors and other electronic components which obtain and process mainly numerical information. Sonar, infrared and laser sensors obtain distance information. Webcams obtain digital images that are represented internally as matrices of red, blue and green (RGB) colour coordinate values. All this numerical values obtained from the environment need a later interpretation in order to provide the knowledge required by the robotic agent in order to carry out a task. Similarly, light wavelengths with specific amplitude are captured by cone cells of human eyes obtaining also stimulus without meaning. However, the information that human beings can describe and remember from what they see is expressed using words, that is qualitatively. The exact process carried out after our eyes perceive light wavelengths and our brain interpret them is quite unknown. However, a real fact in human cognition is that people go beyond the purely perceptual experience to classify things as members of categories and attach linguistic labels to them. As the information provided by all the electronic components incorporated in a robotic agent is numerical, the approaches that first appeared in the literature giving an interpretation of this information followed a mathematical trend. In this thesis, this problem is addressed from the other side, its main aim is to process these numerical data in order to obtain qualitative information as human beings can do. The research work done in this thesis tries to narrow the gap between the acquisition of low level information by robot sensors and the need of obtaining high level or qualitative information for enhancing human-machine communication and for applying logical reasoning processes based on concepts. Moreover, qualitative concepts can be added a meaning by relating them to others. They can be used for reasoning applying qualitative models that have been developed in the last twenty years for describing and interpreting metrical and mathematical concepts such as orientation, distance, velocity, acceleration, and so on. And they can be also understood by human-users both written and read aloud. The first contributions presented are the definition of a method for obtaining fuzzy distance patterns (which include qualitative distances such as ‘near’, far’, ‘very far’ and so on) from the data obtained by any kind of distance sensors incorporated in a mobile robot and the definition of a factor to measure the dissimilarity between those fuzzy patterns. Both have been applied to the integration of the distances obtained by the sonar and laser distance sensors incorporated in a Pioneer 2 dx mobile robot and, as a result, special obstacles have been detected as ‘glass window’, ‘mirror’, and so on. Moreover, the fuzzy distance patterns provided have been also defuzzified in order to obtain a smooth robot speed and used to classify orientation reference systems into ‘open’ (it defines an open space to be explored) or ‘closed’. The second contribution presented is the definition of a model for qualitative image description (QID) by applying the new defined models for qualitative shape and colour description and the topology model by Egenhofer and Al-Taha [1992] and the orientation models by Hernández [1991] and Freksa [1992]. This model can qualitatively describe any kind of digital image and is independent of the image segmentation method used. The QID model have been tested in two scenarios in robotics: (i) the description of digital images captured by the camera of a Pioneer 2 dx mobile robot and (ii) the description of digital images of tile mosaics taken by an industrial camera located on a platform used by a robot arm to assemble tile mosaics. In order to provide a formal and explicit meaning to the qualitative description of the images generated, a Description Logic (DL) based ontology has been designed and presented as the third contribution. Our approach can automatically process any random image and obtain a set of DL-axioms that describe it visually and spatially. And objects included in the images are classified according to the ontology schema using a DL reasoner. Tests have been carried out using digital images captured by a webcam incorporated in a Pioneer 2 dx mobile robot. The images taken correspond to the corridors of a building at University Jaume I and objects with them have been classified into ‘walls’, ‘floor’, ‘office doors’ and ‘fire extinguishers’ under different illumination conditions and from different observer viewpoints. The final contribution is the definition of a similarity measure between qualitative descriptions of shape, colour, topology and orientation. And the integration of those measures into the definition of a general similarity measure between two qualitative descriptions of images. These similarity measures have been applied to: (i) extract objects with similar shapes from the MPEG7 CE Shape-1 library; (ii) assemble tile mosaics by qualitative shape and colour similarity matching; (iii) compare images of tile compositions; and (iv) compare images of natural landmarks in a mobile robot world for their recognition. The contributions made in this thesis are only a small step forward in the direction of enhancing robot knowledge acquisition from the world. And it is also written with the aim of inspiring others in their research, so that bigger contributions can be achieved in the future which can improve the life quality of our society.

Image segmentation is a critical step in building a computer vision algorithm that is able to distinguish between separate objects in an image scene. Image segmentation is based on two fundamentally intertwined components: pixel comparison and pixel grouping. In the pixel comparison step, pixels are determined to be similar or different from each other. In pixel grouping, those pixels which are similar are grouped together to form meaningful regions which can later be processed. This thesis makes original contributions to both of those areas. First, given a Markov Random Field framework, a Stochastic Nested Aggregation (SNA) framework for pixel and region grouping is presented and thoroughly analyzed using a Potts model. This framework is applicable in general to graph partitioning and discrete estimation problems where pairwise energy models are used. Nested aggregation reduces the computational complexity of stochastic algorithms such as Simulated Annealing to order O(N) while at the same time allowing local deterministic approaches such as Iterated Conditional Modes to escape most local minima in order to become a global deterministic optimization method. SNA is further enhanced by the introduction of a Graduated Models strategy which allows an optimization algorithm to converge to the model via several intermediary models. A well-known special case of Graduated Models is the Highest Confidence First algorithm which merges pixels or regions that give the highest global energy decrease. Finally, SNA allows us to use different models at different levels of coarseness. For coarser levels, a mean-based Potts model is introduced in order to compute region-to-region gradients based on the region mean and not edge gradients. Second, we develop a probabilistic framework based on hypothesis testing in order to achieve color constancy in image segmentation. We develop three new shading invariant semi-metrics based on the Dichromatic Reflection Model. An RGB image is transformed into an R'G'B' highlight invariant space to remove any highlight components, and only the component representing color hue is preserved to remove shading effects. This transformation is applied successfully to one of the proposed distance measures. The probabilistic semi-metrics show similar performance to vector angle on images without saturated highlight pixels; however, for saturated regions, as well as very low intensity pixels, the probabilistic distance measures outperform vector angle. Third, for interferometric Synthetic Aperture Radar image processing we apply the Potts model using SNA to the phase unwrapping problem. We devise a new distance measure for identifying phase discontinuities based on the minimum coherence of two adjacent pixels and their phase difference. As a comparison we use the probabilistic cost function of Carballo as a distance measure for our experiments.

Image segmentation is a critical step in building a computer vision algorithm that is able to distinguish between separate objects in an image scene. Image segmentation is based on two fundamentally intertwined components: pixel comparison and pixel grouping. In the pixel comparison step, pixels are determined to be similar or different from each other. In pixel grouping, those pixels which are similar are grouped together to form meaningful regions which can later be processed. This thesis makes original contributions to both of those areas. First, given a Markov Random Field framework, a Stochastic Nested Aggregation (SNA) framework for pixel and region grouping is presented and thoroughly analyzed using a Potts model. This framework is applicable in general to graph partitioning and discrete estimation problems where pairwise energy models are used. Nested aggregation reduces the computational complexity of stochastic algorithms such as Simulated Annealing to order O(N) while at the same time allowing local deterministic approaches such as Iterated Conditional Modes to escape most local minima in order to become a global deterministic optimization method. SNA is further enhanced by the introduction of a Graduated Models strategy which allows an optimization algorithm to converge to the model via several intermediary models. A well-known special case of Graduated Models is the Highest Confidence First algorithm which merges pixels or regions that give the highest global energy decrease. Finally, SNA allows us to use different models at different levels of coarseness. For coarser levels, a mean-based Potts model is introduced in order to compute region-to-region gradients based on the region mean and not edge gradients. Second, we develop a probabilistic framework based on hypothesis testing in order to achieve color constancy in image segmentation. We develop three new shading invariant semi-metrics based on the Dichromatic Reflection Model. An RGB image is transformed into an R'G'B' highlight invariant space to remove any highlight components, and only the component representing color hue is preserved to remove shading effects. This transformation is applied successfully to one of the proposed distance measures. The probabilistic semi-metrics show similar performance to vector angle on images without saturated highlight pixels; however, for saturated regions, as well as very low intensity pixels, the probabilistic distance measures outperform vector angle. Third, for interferometric Synthetic Aperture Radar image processing we apply the Potts model using SNA to the phase unwrapping problem. We devise a new distance measure for identifying phase discontinuities based on the minimum coherence of two adjacent pixels and their phase difference. As a comparison we use the probabilistic cost function of Carballo as a distance measure for our experiments.

Elastography, which is based on applying pressure and estimating the resulting deformation, involves the forward problem to obtain the strain distributions and inverse problem to construct the elastic distributions consistent with the obtained strains on observation points. This thesis focuses on the former problem whose solution is used as an input to the latter problem. The aim is to provide the inverse problem community with accurate strain estimates of a coronary artery vessel wall. In doing so, a new ultrasonic image-based elastography approach is developed. Because the accuracy and quality of the estimated strain fields depend on the resolution level of the ultrasound image and to date best resolution levels obtained in the literature are not enough to clearly see all boundaries of the artery, one of the main goals is to acquire high-resolution coronary vessel wall ultrasound images at different pressures. For this purpose, first an experimental setup is designed to collect radio frequency (RF) signals, and then image formation algorithm is developed to obtain ultrasound images from the collected signals. To segment the noisy ultrasound images formed, a geodesic active contour-based segmentation algorithm with a novel stopping function that includes local phase of the image is developed. Then, region-based information is added to make the segmentation more robust to noise. Finally, elliptical deformable template is applied so that a priori information regarding the shape of the arteries could be taken into account, resulting in more stable and accurate results. The use of this template also implicitly provides boundary point correspondences from which high-resolution, size-independent, non-rigid and local strain fields of the coronary vessel wall are obtained.

The aim of this thesis is to investigate the shape change in hippocampus due to the atrophy in Alzheimer's disease (AD). To this end, specific algorithms and methodologies were developed to segment the hippocampus from structural magnetic resonance (MR) images and model variations in its shape. We use a multi-atlas based segmentation propagation approach for the segmentation of hippocampus which has been shown to obtain accurate parcellation of brain structures. We developed a supervised method to build a population specific atlas database, by propagating the parcellations from a smaller generic atlas database. Well segmented images are inspected and added to the set of atlases, such that the segmentation capability of the atlas set may be enhanced. The population specific atlases are evaluated in terms of the agreement among the propagated labels when segmenting new cases. Compared with using generic atlases, the population specific atlases obtain a higher agreement when dealing with images from the target population. Atlas selection is used to improve segmentation accuracy. In addition to the conventional selection by image similarity ranking, atlas selection based on maximum marginal relevance (MMR) re-ranking and least angle regression (LAR) sequence are developed for atlas selection. By taking the redundancy among atlases into consideration, diversity criteria are shown to be more efficient in atlas selection which is applicable in the situation where the number of atlases to be fused is limited by the computational resources. Given the segmented hippocampal volumes, statistical shape models (SSMs) of hippocampi are built on the samples to model the shape variation among the population. The correspondence across the training samples of hippocampi is established by a groupwise optimization of the parameterized shape surfaces. The spherical parameterization of the hippocampal surfaces are flatten to facilitate the reparameterization and interpolation. The reparameterization is regularized by viscous fluid, which is solved by a fast implementation based on discrete sine transform. In order to use the hippocampal SSM to describe the shape of an unseen hippocampal surface, we developed a shape parameter estimator based on the expectationmaximization iterative closest points (EM-ICP) algorithm. A symmetric data term is included to achieve the inverse consistency of the transformation between the model and the shape, which gives more accurate reconstruction of the shape from the model. The shape prior modeled by the SSM is used in the maximum a posteriori estimation of the shape parameters, which is shown to enforce the smoothness and avoid the effect of over-fitting. In the study of the hippocampus in AD, we use the SSM to model the hippocampal shape change between the healthy control subjects and patients diagnosed with AD. We identify the regions affected by the atrophy in AD by assessing the spatial difference between the control and AD groups at each corresponding landmark. Localized shape analysis is performed on the regions exhibiting significant inter-group difference, which is shown to improve the discrimination ability of the principal component analysis (PCA) based SSM. The principal components describing the localized shape variability among the population are also shown to display stronger correlation with the decline of episodic memory scores linked to the pathology of hippocampus in AD.

Graphics processing units (GPUs) are used today in a wide range of applications, mainly because they can dramatically accelerate parallel computing, are affordable and energy efficient. In the field of medical imaging, GPUs are in some cases crucial for enabling practical use of computationally demanding algorithms. This review presents the past and present work on GPU accelerated medical image processing, and is meant to serve as an overview and introduction to existing GPU implementations. The review covers GPU acceleration of basic image processing operations (filtering, interpolation, histogram estimation and distance transforms), the most commonly used algorithms in medical imaging (image registration, image segmentation and image denoising) and algorithms that are specific to individual modalities (CT, PET, SPECT, MRI, fMRI, DTI, ultrasound, optical imaging and microscopy). The review ends by highlighting some future possibilities and challenges.