The research in the thesis investigates the use of minimal path techniques to track and detect cracks, modeled as curves, in critical infrastructure like pavements and bridges. We developed a novel minimal path algorithm to detect curves with complex topology that may have both closed cycles and open sections using an arbitrary point on the curve as the sole input. Specically, we applied the novel algorithm to three problems: semi-automatic crack detection, detection of continuous cracks for crack sealing applications and detection of crack growth in structures like bridges. The current state of the art minimal path techniques only work with prior knowledge of either both terminal points or one terminal point plus total length of the curve. For curves with multiple branches, all terminal points need to be known. Therefore, we developed a new algorithm that detects curves and relaxes the necessary user input to one arbitrary point on the curve. The document presents the systematic development of this algorithm in three stages. First, an algorithm that can detect open curves with branches was formulated. Then this algorithm was modied to detect curves that also have closed cycles. Finally, a robust curve detection algorithm was devised that can increase the accuracy of curve detection. The algorithm was applied to crack images and the results of crack detection were validated against the ground truth. In addition, the algorithm was also used to detect features like catheter tube and optical nerves in medical images. The results demonstrate that the algorithm is able to accurately detect objects that can be modeled as open curves.

Extracting surfaces, representing boundaries of objects of interest, from volumetric images, has important applications in various scientific domains, from medicine to geology. In this thesis, I introduce novel mathematical, computational, and algorithmic machinery for extraction of sheet-like surfaces (with boundary), whose boundary is unknown a-priori, a particularly important case in applications that has no convenient methods. This case of a surface with boundaries has applications in extracting faults (among other geological structures) from seismic images in geological applications. Another application domain is in the extraction of structures in the lung from computed tomography (CT) images. Although many methods have been developed in computer vision for extraction of surfaces, including level sets, convex optimization approaches, and graph cut methods, none of these methods appear to be applicable to the case of surfaces with boundary. The novel methods for surface extraction, derived in this thesis, are built on the theory of Minimal Paths, which has been used primarily to extract curves in noisy or corrupted images and have had wide applicability in 2D computer vision. This thesis extends such methods to surfaces, and it is based on novel observations that surfaces can be determined by extracting topological structures from the solution of the eikonal partial differential equation (PDE), which is the basis of Minimal Path theory. Although topological structures are known to be difficult to extract from images, which are both noisy and discrete, this thesis builds robust methods based on Morse theory and computational topology to address such issues. The algorithms have run-time complexity O(NlogN), less complex than existing approaches. The thesis details the algorithms, theory, and shows an extensive experimental evaluation on seismic images and medical images. Experiments show out-performance in accuracy, computational speed, and user convenience compared with related state-of-the-art methods. Lastly, the thesis shows the methodology developed for the particular case of surfaces with boundary extends to surfaces without boundary and also surfaces with different topologies, such as cylindrical surfaces, both important cases for many applications in medical image analysis.

Problems involving the minimization of functionals date back to antiquity. The mathematics of the calculus of variations has provided a framework for the analytical solution of a limited class of such problems. This paper describes a numerical approximation technique for obtaining machine solutions to minimal path problems. It is shown that this technique is applicable not only to the common case of finding geodesics on parameterized surfaces in R3, but also to the general case of finding minimal functionals on hypersurfaces in Rn associated with an arbitrary metric.

To aid diagnosis, treatment planning, and patient education, clinicians require tools to anal- yse and explore the increasingly large three-dimensional (3-D) datasets generated by modern medical scanners. Direct volume rendering is one such tool finding favour with radiologists and surgeons for its photorealistic representation. More recently, volume illustration — or non-photorealistic rendering (NPR) — has begun to move beyond the mere depiction of data, borrowing concepts from illustrators to visually enhance desired information and suppress un- wanted clutter. Direct volume rendering generates images by accumulating pixel values along rays cast into a 3-D image. Transfer functions allow users to interactively assign material properties such as colour and opacity (a process known as classification). To achieve real-time framerates, the rendering must be accelerated using a technique such as 3-D texture mapping on commod- ity graphics processing units (GPUs). Unfortunately, current methods do not allow users to intuitively enhance regions of interest or suppress occluding structures. Furthermore, addi- tional scalar images describing clinically relevant measures have not been integrated into the direct rendering method. These tasks are essential for the effective exploration, analysis, and presentation of 3-D images. This body of work seeks to address the aforementioned limitations. First, to facilitate the research program, a flexible architecture for prototyping volume illustration methods is pro- posed. This program unifies a number of existing techniques into a single framework based on 3-D texture mapping, while also providing for the rapid experimentation of novel methods. Next, the prototyping environment is employed to improve an existing method—called tagged volume rendering — which restricts transfer functions to given spatial regions using a number of binary segmentations (tags). An efficient method for implementing binary tagged volume rendering is presented, along with various technical considerations for improving the classifi- cation. Finally, the concept of greyscale tags is proposed, leading to a number of novel volume visualisation techniques including position modulated classification and dynamic exploration. The novel methods proposed in this work are generic and can be employed to solve a wide range of problems. However, to demonstrate their usefulness, they are applied to a specific case study. Ischaemic heart disease, caused by narrowed coronary arteries, is a leading healthconcern in many countries including Australia. Computed tomography angiography (CTA) is an imaging modality which has the potential to allow clinicians to visualise diseased coronary arteries in their natural 3-D environment. To apply tagged volume rendering for this case study, an active contour method and minimal path extraction technique are proposed to segment the heart and arteries respectively. The resultant images provide new insight and possibilities for diagnosing and treating ischaemic heart disease.

Le Doppler ultrasensible est une nouvelle technique d'imagerie ultrasonore permettant d'observer les flux sanguins avec une résolution très fine et sans agent de contraste. Appliquée à l'imagerie microvasculaire cérébrale des rongeurs, cette méthode produit de très fines cartes vasculaires 3D du cerveau à haute résolution spatiale. Ces réseaux vasculaires contiennent des structures tubulaires caractéristiques qui pourraient servir de points de repère pour localiser la position de la sonde ultrasonore et tirer parti des avantages pratiques des appareils à ultrason. Ainsi, nous avons développé un premier système de neuronavigation chez les rongeurs basé sur le recalage automatique d'images cérébrales. En utilisant des méthodes d’extraction de chemins minimaux, nous avons développé une nouvelle méthode isotrope de segmentation pour l’analyse géométrique des réseaux vasculaires en 3D. Cette méthode a été appliquée à la quantification des réseaux vasculaires et a permis le développement d'algorithmes de recalage de nuages de points pour le suivi temporel de tumeurs. / Ultrasensitive Doppler is a new ultrasound imaging technique allowing the observation of blood flows with a very fine resolution and no contrast agent. Applied to cerebral microvascular imaging in rodents, this method produces very fine vascular 3D maps of the brain at high spatial resolution. These vascular networks contain characteristic tubular structures that could be used as landmarks to localize the position of the ultrasonic probe and take advantage of the easy-to-use properties of ultrasound devices such as low cost and portability. Thus, we developed a first neuronavigation system in rodents based on automatic registration of brain images. Using minimal path extraction methods, we developed a new isotropic segmentation framework for 3D geometric analysis of vascular networks (extraction of centrelines, diameters, curvatures, bifurcations). This framework was applied to quantify brain and tumor vascular networks, and finally leads to the development of point cloud registration algorithms for temporal monitoring of tumors.

Dans les domaines de l’imagerie médicale et de la vision par ordinateur, la segmentation joue un rôle crucial dans le but d’extraire les composantes intéressantes d’une image ou d’une séquence d’images. Elle est à l’intermédiaire entre le traitement d’images de bas niveau et les applications cliniques et celles de la vision par ordinateur de haut niveau. Ces applications de haut niveau peuvent inclure le diagnostic, la planification de la thérapie, la détection et la reconnaissance d'objet, etc. Parmi les méthodes de segmentation existantes, les courbes géodésiques minimales possèdent des avantages théoriques et pratiques importants tels que le minimum global de l’énergie géodésique et la méthode bien connue de Fast Marching pour obtenir une solution numérique. Dans cette thèse, nous nous concentrons sur les méthodes géodésiques basées sur l’équation aux dérivées partielles, l’équation Eikonale, afin d’étudier des méthodes précises, rapides et robustes, pour l’extraction de structures tubulaires et la segmentation d’image, en développant diverses métriques géodésiques locales pour des applications cliniques et la segmentation d’images en général. / In the fields of medical imaging and computer vision, segmentation plays a crucial role with the goal of separating the interesting components from one image or a sequence of image frames. It bridges the gaps between the low-level image processing and high level clinical and computer vision applications. Among the existing segmentation methods, minimal geodesics have important theoretical and practical advantages such as the global minimum of the geodesic energy and the well-established fast marching method for numerical solution. In this thesis, we focus on the Eikonal partial differential equation based geodesic methods to investigate accurate, fast and robust tubular structure extraction and image segmentation methods, by developing various local geodesic metrics for types of clinical and segmentation tasks. This thesis aims to applying different geodesic metrics based on the Eikonal framework to solve different image segmentation problems especially for tubularity segmentation and region-based active contours models, by making use of more information from the image feature and prior clinical knowledges. The designed geodesic metrics basically take advantages of the geodesic orientation or anisotropy, the image feature consistency, the geodesic curvature and the geodesic asymmetry property to deal with various difficulties suffered by the classical minimal geodesic models and the active contours models. The main contributions of this thesis lie at the deep study of the various geodesic metrics and their applications in medical imaging and image segmentation. Experiments on medical images and nature images show the effectiveness of the presented contributions.