Vasculature is known to be of key biological significance, especially in the study of tumors. As such, considerable effort has been focused on the automated segmentation of vasculature in medical and pre-clinical images. The majority of vascular segmentation methods focus on bloodpool labeling methods; however, particularly, in the study of tumors, it is of particular interest to be able to visualize both the perfused and the non-perfused vasculature. Imaging vasculature by highlighting the endothelium provides a way to separate the morphology of vasculature from the potentially confounding factor of perfusion. Here, we present a method for the segmentation of tumor vasculature in 3D fluorescence microscopic images using signals from the endothelial and surrounding cells. We show that our method can provide complete and semantically meaningful segmentations of complex vasculature using a supervoxel-Markov random field approach. We show that in terms of extracting meaningful segmentations of the vasculature, our method outperforms both state-of-the-art method, specific to these data, as well as more classical vasculature segmentation methods.

Rectal tumour segmentation in dynamic contrast-enhanced MRI (DCE-MRI) is a challenging task, and an automated and consistent method would be highly desirable to improve the modelling and prediction of patient outcomes from tissue contrast enhancement characteristics - particularly in routine clinical practice. A framework is developed to automate DCE-MRI tumour segmentation, by introducing: perfusion-supervoxels to over-segment and classify DCE-MRI volumes using the dynamic contrast enhancement characteristics; and the pieces-of-parts graphical model, which adds global (anatomic) constraints that further refine the supervoxel components that comprise the tumour. The framework was evaluated on 23 DCE-MRI scans of patients with rectal adenocarcinomas, and achieved a voxelwise area-under the receiver operating characteristic curve (AUC) of 0.97 compared to expert delineations. Creating a binary tumour segmentation, 21 of the 23 cases were segmented correctly with a median Dice similarity coefficient (DSC) of 0.63, which is close to the inter-rater variability of this challenging task. A second study is also included to demonstrate the method's generalisability and achieved a DSC of 0.71. The framework achieves promising results for the underexplored area of rectal tumour segmentation in DCE-MRI, and the methods have potential to be applied to other DCE-MRI and supervoxel segmentation problems.

The goal of this work is to reliably and accurately localize anatomical landmarks in 3-D computed tomography scans, particularly for the deformable registration of whole-body scans, which show huge variation in posture, and the spatial distribution of anatomical features. Parts-based graphical models (GM) have shown attractive properties for this task because they capture naturally anatomical relationships between landmarks. Unfortunately, standard GMs are learned from manually annotated training images and the quantity of landmarks is limited by the high cost of expert annotation. We propose a novel method that automatically learns new corresponding landmarks from a database of 3-D whole-body CT scans, using a limited initial set of expert-labeled ground-truth landmarks. The newly learned landmarks, called B-landmarks, are used to build enriched GMs. We compare our method of deformable registration based on such GM landmarks to a conventional deformable registration method and to a "baseline" state-of-the-art GM. The results show our method finds new relevant anatomical correspondences and improves by up to 35% the matching accuracy of highly variable skeletal and soft-tissue landmarks of clinical interest.

Deformable image registration is an important tool in medical image analysis. In the case of lung computed tomography (CT) registration there are three major challenges: large motion of small features, sliding motions between organs, and changing image contrast due to compression. Recently, Markov random field (MRF)-based discrete optimization strategies have been proposed to overcome problems involved with continuous optimization for registration, in particular its susceptibility to local minima. However, to date the simplifications made to obtain tractable computational complexity reduced the registration accuracy. We address these challenges and preserve the potentially higher quality of discrete approaches with three novel contributions. First, we use an image-derived minimum spanning tree as a simplified graph structure, which copes well with the complex sliding motion and allows us to find the global optimum very efficiently. Second, a stochastic sampling approach for the similarity cost between images is introduced within a symmetric, diffeomorphic B-spline transformation model with diffusion regularization. The complexity is reduced by orders of magnitude and enables the minimization of much larger label spaces. In addition to the geometric transform labels, hyper-labels are introduced, which represent local intensity variations in this task, and allow for the direct estimation of lung ventilation. We validate the improvements in accuracy and performance on exhale-inhale CT volume pairs using a large number of expert landmarks.

Recently, Kadir and Brady proposed a method for estimating probability density functions (PDFs) for digital signals which they call the Nonparametric (NP) Windows method. The method involves constructing a continuous space representation of the discrete space and sampled signal by using a suitable interpolation method. NP Windows requires only a small number of observed signal samples to estimate the PDF and is completely data driven. In this short paper, we first develop analytical formulae to obtain the NP Windows PDF estimates for 1D, 2D, and 3D signals, for different interpolation methods. We then show that the original procedure to calculate the PDF estimate can be significantly simplified and made computationally more efficient by a judicious choice of the frame of reference. We have also outlined specific algorithmic details of the procedures enabling quick implementation. Our reformulation of the original concept has directly demonstrated a close link between the NP Windows method and the Kernel Density Estimator.

Positron emission tomography (PET) is a molecular imaging technique which provides important functional information about the human body. However, thoracic PET images are often substantially degraded by respiratory motion, which adversely impacts on subsequent diagnosis. In this paper, a motion correction and attenuation correction method is proposed to correct for motion in respiratory gated PET images and to yield an accurate distribution of the radioactivity concentration. Experimental results show that this method can effectively correct for motion and improve PET image quality. The method is able to provide improved diagnostic information without increasing the acquisition time or the radiation burden.

We propose a method for reliably and accurately identifying anatomical landmarks in 3D CT volumes based on dense matching of parts-based graphical models. Such a system can be used to establish reliable correspondences in medical images which can be useful on their own or as part of more complex processing e.g. atlas building. We propose and investigate novel methods for efficiently optimizing parameters of appearance models for landmark localization in 3D images. We also investigate the trade-off between the number of model parameters and registration accuracy. We present results for the localization of 22 landmarks in clinical 3D CT volumes of cancer patients and optimization of part-specific patch scales. Over-fitting is likely due to an intrinsically high variability of the data and a limited labeled training and test set, here 83 scans, so we employ a rigorous bootstrap analysis to validate the results. The average mean and maximum registration error over all landmarks is reduced by 31% and 25% for the optimized model, compared to an empirically determined baseline. Additionally, we show a significantly improved performance over standard methods as the number of free parameters increases from an isotropic patch scale shared by all parts, to specific anisotropic patch scales learnt for each part in the model.

In multidisciplinary team (MDT) meetings for colorectal and liver cancer, each patient case is reviewed while evidence, including digital image scans such as MRI and PET/CT, is presented by clinicians. Currently these images are projected onto a wall, limiting clinician interaction. While multi-touch and interactive tabletops have been used to enhance collaboration in various scenarios, some aspects such as image quality and touch resolution need to be evaluated in this particular scenario. In this paper we present the results of work conducted to test the suitability of using a DiamondTouch tabletop, a multi-touch and multi-user surface, in MDT meetings to enhance clinician interaction.

Tomographic reconstruction of fluorescence optical projection tomography (OPT) data is usually performed using the standard filtered back projection (FBP) algorithm. However, there are several physical aspects of fluorescence OPT that pose major challenges for the FBP algorithm. These include blurring, and the fact that for an isotropically emitting point source (or fluorophore), the power received by an objective aperture decreases with the inverse square of the distance to the source. These two effects are shown to result in qualitative and quantitative inaccuracies in fluorescence OPT reconstructions obtained using standard FBP. A model of image formation is developed which includes the effects of isotropic emission and blurring. The model is used to calculate a probabilistic system matrix for use in the maximum likelihood expectation maximisation algorithm, which leads to reconstructions that are both qualitatively superior and quantitatively correct.

There are various issues that limit the development and deployment of new software solutions in cancer image analysis research. In this paper we discuss some of these and propose a framework design based on cloud computing concepts, Microsoft technologies, existing middleware and imaging toolkits. Furthermore, we address some of these issues by introducing collaborative visual tools for visual input data and multi-user interactions.

Textures within real images vary in brightness, contrast, scale and skew as imaging conditions change. To enable recognition of textures in real images, it is necessary to employ a similarity measure which is invariant to these properties. Furthermore, since textures often appear on undulating surfaces, such invariances must necessarily be local rather than global. Despite these requirements, it is only relatively recently that texture recognition algorithms with local scale and affine invariance properties have begun to be reported. Typically, they comprise detecting feature points followed by geometric normalization prior to description. We describe a method based on invariant combinations of linear filters. Unlike previous methods, we introduce a novel family of filters, which provide scale invariance, resulting in a texture description invariant to local changes in orientation, contrast and scale and robust to local skew. Significantly, the family of filters enable local scale invariants to be defined without using a scale selection principle or a large number of filters. A texture discrimination method based on the A2 similarity measure applied to histograms derived from our filter responses outperforms existing methods for retrieval and classification results for both the Brodatz textures and the UIUC database, which has been designed to require local invariance.

In recent years, a significant number of developments across a broad range of disciplines have allowed researchers and clinicians to start to build up a picture of cancer development. In this paper we report upon the development of a prototype of a secure distributed infrastructure that links imaging data from pathology and radiology. The intention is that a fully-developed system will be capable of supporting studies that will examine whether prognostic and diagnostic features which are apparent in histopathological sections and clinical scans are related. Further, these studies will consider whether these features can be meaningfully linked into a diagnostic or predictive profile. The project in which the prototype is being developed naturally involves a large degree of cooperation across various disciplines. The focus of this paper is primarily on the development of the underlying prototype infrastructure.

This paper presents a robust algorithm for the comparison of mammogram pairs. Salient regions are extracted in a topographic way. An integral invariant representation of shape, in combination with area and distance measures, are used for establishing their correspondences. The experimental results demonstrate that the algorithm can provide a useful tool for a ComputerAided Diagnosis system in mammography.

We address the problem of estimating statistics in regions of interest (ROIs) containing both whole and partial pixels. Such ROIs arise frequently in vision problems such as segmentation and registration. For example, even where the control points of an ROI, say the vertices of a polygon, are forcibly aligned with the pixel grid, the connecting edges will rarely do so. In medical image analysis, for instance, this can be a cause of significant error. More generally, any cost function that includes statistics estimated from the image will often exhibit irregularities due to such partial pixels. Our proposed solution addresses this problem by correctly accounting for the partial pixel area. Moreover, the method has no arbitrary parameters such as bin widths or kernel sizes. It implicitly addresses the issue of independence and gives rise to continuous density estimates whose quality is, in principle at least, independent of the number of pixels in the ROI. We present results to compare our proposed method with conventional techniques such as weighted histograms and Parzen windowing.

We consider the registration of successive images over a cardiac ultrasound sequence. A key challenge is speckle. We suggest that, for practical purposes, speckle should be regarded as the sum of two components: the useful, temporally correlated speckle pattern; and speckle noise, which is the unpredictable change in the speckle pattern. We also propose that an increase in robustness can be obtained by modelling the spatial correlation of speckle noise. A model is introduced that captures the correlation of speckle noise, by separating the noise into sub-bands and measuring variance within each sub-band and covariance between bands. In order to gain contrast invariance, we express similarity between sub-bands in terms of local phase only, and describe a phase-based extension of the Demons algorithm. The parameters estimated from the noise model significantly enhance the already good registration performance of the phase-demons algorithm. 1

The NCRI Informatics Initiative is overseeing the implementation of an informatics framework for the UK cancer research community. The framework advocates an integrated multidisciplinary method of working between scientific and medical communities. Key to this process is community adoption of high quality acquisition, storage, sharing and integration of diverse data elements to improve knowledge of the causes, prevention and treatment of cancer. The integration of the complex data and meta-data used by these multiple communities is a significant challenge and there are technical, resource-based and sociological issues to be addressed. In this paper we review progress aimed at establishing the framework and outline key challenges in ultra large scale integration of biomedical computing systems.

In this paper, we present a two-stage algorithm for mammogram registration, the geometrical alignment of mammogram sequences. The rationale behind this paper stems from the intrinsic difficulties in comparing mammogram sequences. Mammogram comparison is a valuable tool in national breast screening programs as well as in frequent monitoring and hormone replacement therapy (HRT). The method presented in this paper aims to improve mammogram comparison by estimating the underlying geometric transformation for any mammogram sequence. It takes into consideration the various temporal changes that may occur between successive scans of the same woman and is designed to overcome the inconsistencies of mammogram image formation.

We present a fully Bayesian approach to modeling in functional magnetic resonance imaging (FMRI), incorporating spatio-temporal noise modeling and haemodynamic response function (HRF) modeling. A fully Bayesian approach allows for the uncertainties in the noise and signal modeling to be incorporated together to provide full posterior distributions of the HRF parameters. The noise modeling is achieved via a nonseparable space-time vector autoregressive process. Previous FMRI noise models have either been purely temporal, separable or modeling deterministic trends. The specific form of the noise process is determined using model selection techniques. Notably, this results in the need for a spatially nonstationary and temporally stationary spatial component. Within the same full model, we also investigate the variation of the HRF in different areas of the activation, and for different experimental stimuli. We propose a novel HRF model made up of half-cosines, which allows distinct combinations of parameters to represent characteristics of interest. In addition, to adaptively avoid over-fitting we propose the use of automatic relevance determination priors to force certain parameters in the model to zero with high precision if there is no evidence to support them in the data. We apply the model to three datasets and observe matter-type dependence of the spatial and temporal noise, and a negative correlation between activation height and HRF time to main peak (although we suggest that this apparent correlation may be due to a number of different effects).

We present a novel non-parametric unsupervised segmentation algorithm based on region competition (Zhu and Yuille, 1996); but implemented within a level sets framework (Osher and Sethian, 1988). The key novelty of the algorithm is that it can solve N /spl ges/ 2 class segmentation problems using just one embedded surface; this is achieved by controlling the merging and splitting behaviour of the level sets according to a minimum description length (MDL) (Leclerc (1989) and Rissanen (1985)) cost function. This is in contrast to N class region-based level set segmentation methods to date which operate by evolving multiple coupled embedded surfaces in parallel (Chan et al., 2002). Furthermore, it operates in an unsupervised manner; it is necessary neither to specify the value of N nor the class models a-priori. We argue that the level sets methodology provides a more convenient framework for the implementation of the region competition algorithm, which is conventionally implemented using region membership arrays due to the lack of a intrinsic curve representation. Finally, we generalise the Gaussian region model used in standard region competition to the non-parametric case. The region boundary motion and merge equations become simple expressions containing cross-entropy and entropy terms.

Three-dimensional (3-D) ultrasound imaging of the breast enables better assessment of diseases than conventional two-dimensional (2-D) imaging. Free-hand techniques are often used for generating 3-D data from a sequence of 2-D slice images. However, the breast deforms substantially during scanning because it is composed primarily of soft tissue. This often causes tissue mis-registration in spatial compounding of multiple scan sweeps. To overcome this problem, in this paper, instead of introducing additional constraints on scanning conditions, we use image processing techniques. We present a fully automatic algorithm for 3-D nonlinear registration of free-hand ultrasound data. It uses a block matching scheme and local statistics to estimate local tissue deformation. A Bayesian regularization method is applied to the sample displacement field. The final deformation field is obtained by fitting a B-spline approximating mesh to the sample displacement field. Registration accuracy is evaluated using phantom data and similar registration errors are achieved with (0.19 mm) and without (0.16 mm) gaps in the data. Experimental results show that registration is crucial in spatial compounding of different sweeps. The execution time of the method on moderate hardware is sufficiently fast for fairly large research studies.

Displayed ultrasound (US) B-mode images often exhibit tissue intensity inhomogeneities dominated by nonuniform beam attenuation within the body. This is a major problem for intensity-based, automatic segmentation of video-intensity images because conventional threshold-based or intensity-statistic-based approaches do not work well in the presence of such image distortions. Time gain compensation (TGC) is typically used in standard US machines in an attempt to overcome this. However this compensation method is position-dependent which means that different tissues in the same TGC time-range (or corresponding depth range) will be, incorrectly, compensated by the same amount. Compensation should really be tissue-type dependent but automating this step is difficult. The main contribution of this paper is to develop a method for simultaneous estimation of video-intensity inhomogeities and segmentation of US image tissue regions. The method uses a combination of the maximum a posteriori (MAP) and Markov random field (MRF) methods to estimate the US image distortion field assuming it follows a multiplicative model while at the same time labeling image regions based on the corrected intensity statistics. The MAP step is used to estimate the intensity model parameters while the MRF step provides a novel way of incorporating the distributions of image tissue classes as a spatial smoothness constraint. We explain how this multiplicative model can be related to the ultrasonic physics of image formation to justify our approach. Experiments are presented on synthetic images and a gelatin phantom to evaluate quantitatively the accuracy of the method. We also discuss qualitatively the application of the method to clinical breast and cardiac US images. Limitations of the method and potential clinical applications are outlined in the conclusion.

The finite mixture (FM) model is the most commonly used model for statistical segmentation of brain magnetic resonance (MR) images because of its simple mathematical form and the piecewise constant nature of ideal brain MR images. However, being a histogram-based model, the FM has an intrinsic limitation--no spatial information is taken into account. This causes the FM model to work only on well-defined images with low levels of noise; unfortunately, this is often not the the case due to artifacts such as partial volume effect and bias field distortion. Under these conditions, FM model-based methods produce unreliable results. In this paper, we propose a novel hidden Markov random field (HMRF) model, which is a stochastic process generated by a MRF whose state sequence cannot be observed directly but which can be indirectly estimated through observations. Mathematically, it can be shown that the FM model is a degenerate version of the HMRF model. The advantage of the HMRF model derives from the way in which the spatial information is encoded through the mutual influences of neighboring sites. Although MRF modeling has been employed in MR image segmentation by other researchers, most reported methods are limited to using MRF as a general prior in an FM model-based approach. To fit the HMRF model, an EM algorithm is used. We show that by incorporating both the HMRF model and the EM algorithm into a HMRF-EM framework, an accurate and robust segmentation can be achieved. More importantly, the HMRF-EM framework can easily be combined with other techniques. As an example, we show how the bias field correction algorithm of Guillemaud and Brady (1997) can be incorporated into this framework to achieve a three-dimensional fully automated approach for brain MR image segmentation.

We propose a method to solve the general problem of road tracking and 3D-shape reconstruction for a smart vehicle. The method assumes that the road boundaries are parallel and that the width of the road is constant. We then detect and track the road region in the image using active contour models subject to a parallelism constraint. The system then generates a 3D-road model from a single image. We evaluate the effectiveness of the method by applying to real road scenes comprising more than 2000 images.

Microcalcifications can be one of the earliest signs of breast cancer. Unfortunately, their appearance in mammograms can be mimicked by dust and dirt entering the imaging process and this has been shown previously to lead to false positives. We use a model of the imaging process and, in particular, the blurring functions inherent within it to detect the film-screen artifacts caused by dust and dirt and, thus, reduce false-positives. A crucial facet of the work is the choice of the correct image representation upon which to perform the image processing. After extensive testing, our algorithm has identified no microcalcifications as being artifacts and has an artifact detection rate of approaching 96%.

We develop a method to find correspondences between a Cranio-Caudal (CC) and a Medio-Lateral Oblique (MLO) X-ray image of the same breast. Matching between such pairs of images is considered essential by radiologists for more reliable diagnosis of early breast cancer. The two images are taken while the breast is compressed between the cassette and plate of the X-ray machine, but, almost always, to a different extent in each direction. The deformations of the breast caused by the different compressions in the different directions causes corresponding points to appear far from the straight "epipolar lines" familiar from binocular stereo vision. The method developed in this paper calculates the line in a MLO image corresponding to a point in the CC image through simulation of the deformation and the projection of a 3D line (curve) corresponding to the point. Experiments using actual images show that the method gives good predictions which can be used to find exact correspondences between points in the two images.

Modern manufacturing requires flexible and autonomous systems for materials handling and transportation. For such applications, a succession of implemented mobile robots has been developed over the past seven years. In this paper, the questions of software and hardware architecture are discussed in light of experience gained so far. Also discussed are mobile robot projects that seem realistic over the next five to ten years.

We devise enhancement algorithms for far infrared images based upon a model of an idealized far infrared image being piecewise-constant. We then apply two known enhancement algorithms: median filtering and spatial homomorphic filtering, and then extend the model to develop spatio-temporal homomorphic filtering. The algorithms have been applied to several image sequences and work well, showing significant image enhancement.

We propose a modification of Wells et al. technique for bias field estimation and segmentation of magnetic resonance (MR) images. We show that replacing the class other, which includes all tissue not modeled explicitly by Gaussians with small variance, by a uniform probability density, and amending the expectation-maximization (EM) algorithm appropriately, gives significantly better results. We next consider the estimation and filtering of high-frequency information in MR images, comprising noise, intertissue boundaries, and within tissue microstructures. We conclude that post-filtering is preferable to the prefiltering that has been proposed previously. We observe that the performance of any segmentation algorithm, in particular that of Wells et al. (and our refinements of it) is affected substantially by the number and selection of the tissue classes that are modeled explicitly, the corresponding defining parameters and, critically, the spatial distribution of tissues in the image. We present an initial exploration to choose automatically the number of classes and the associated parameters that give the best output. This requires us to define what is meant by "best output" and for this we propose the application of minimum entropy. The methods developed have been implemented and are illustrated throughout on simulated and real data (brain and breast MR).

We propose a probabilistic approach to the problem of global path planning with uncertainty for mobile robots in a dynamic manufacturing environment. To model the changing environment, we use a topological graph weighted by scalar cost functions. The cost functions consist of two elements: a deterministic cost for the known part of the robot's environment, and an uncertainty cost for the unknown part of the environment. Statistical models are built to quantify the unknown part of the environment, forming uncertainty costs for handling unexpected events. These uncertainty costs are dynamically updated by available sensor data when the mobile robot moves around. An optimal path (suboptimal in practice) is then found from the weighted topological graph using dynamic programming.

The vast majority of corner and edge detectors measure image intensity gradients in order to estimate the positions and strengths of features. However, many of the most popular intensity gradient estimators are inherently and significantly anisotropic. In spite of this, few algorithms take the anisotropy into account, and so the set of features uncovered is typically sensitive to rotations of the image, compromising recognition, matching (e.g. stereo), and tracking. We introduce an effective technique for removing unwanted anisotropies from analytical gradient estimates, by measuring local intensity gradients in four directions rather than the more traditional two. In experiments using real image data, our algorithm reduces the gradient anisotropy associated with conventional analytical gradient estimates by up to 85%, yielding more consistent feature topologies.

A significant drawback of the symmetry set evaluation for a 2D shape using the wave diffusion process1 is its slow execution time caused, in large part, by the diffusion step. We first recall the need for a diffusion process. A parallel implementation of the wave diffusion algorithm on a transputer network is presented. A faster alternative approach to detect the symmetry set, which we call the normal transform, and which is similar to the wave propagation mechanism, is described.

We describe a number of advances in the analysis of road scenes when the scene contains multiple moving objects and is observed by a single nonsteerable camera mounted on the front of a vehicle. Our structure from motion approach to scene segmentation derives front the observed motions of independently moving objects and requires no prior knowledge. We describe a hierarchy of camera models for the analysis of the scene, the simpler models handle degeneracies that occur in the more complex models. The major technical contribution is the recursive computation of feature clusters, which are fed forward over time. This closed loop feature tracking generates extended feature trajectories which significantly improve the discriminating power of scene segmentation.>

For many image processing tasks, the wavelet transform is used to represent an image by oriented spatial-frequency (scale-space) channels in which some properties of the image are better represented than in image space. The spatial behaviour in each channel, and the relationships between the channels are critically important for subsequent processing. In this paper, a 2D local energy and phase representation of a wavelet transform is presented. Based on this representation, we note that in general a wavelet transform is coupled with the phase component of the analysing wavelet associated with that scale and orientation. Consequently, commonly used features, such as squaring, and half- or full-wave rectification of a wavelet transform also depend on this phase component which not only causes unnecessary spatial variation of features at each scale but also makes it more difficult to associate features meaningfully across scales. Instead, the 2D local energy of a wavelet transform is proposed as a local feature for image texture segmentation. The advantages of using this local energy feature are that the feature is not only immune to spatial variations caused by the phase component of the analysing wavelet but can be related from one scale to another. The success of the approach is demonstrated by experimental results for both real infrared line scan (IRLS) aerial images and Brodatz images.

A modular architecture used for intelligent control of mobile robots has been developed at Oxford University over the last few years. This architecture takes the form of multiple sensing and control layers, based on "locally intelligent control agents" (LICAs). Central to such a design is the concept of "communication sequential processes" (CSP). This paper describes briefly the LICA-based control architecture for an Oxford mobile robot. Two typical operations of sonar and active vision systems have been given to demonstrate its flexibility and efficiency. Such a modular and unified system has been licensed commercially and allows for the integration of all kinds of sensors and actuators to meet future needs.

Abstract We present in the paper a heterogeneous machine, PARADOX, designed in Oxford for real time image processing. PARADOX consists of a Datacube pipelined image processor, a configurable network of 32 T800 transputers, a Sun‐4 workstation and a special‐purpose interface connecting the Datacube to the transputer network. Its programming style is a mixture of MIMD paradigm employing a processor farm and interrupts to control the image processing pipeline in the Datacube via a VME bus. A number of vision algorithms including the Canny edge detector have been implemented on this machine at video rate. In particular, a 3D vision algorithm, Droid, has been implemented to provide navigation data for an autonomous guided vehicle.

This paper describes a system for detecting and tracking moving objects in a moving world. The feature-based optic flow field is segmented into clusters with affine internal motion which are tracked over time. The system runs in real-time, and is accurate and reliable.>

A new corner detection algorithm has been developed based on the observation of surface curvature. The algorithm utilizes a linear interpolation scheme for intermediate pixel addressing in the differentiation step, which results in improved accuracy of corner localisation and reduced computational complexity. Noise is reduced by a combination of Gaussian convolution, non-maximum suppression and false corner response suppression. The corner finder is applied to computing stereo disparities and structures from motion. It is implemented on a hybrid parallel processor PARADOX with a performance of 14 frames per second.>

The design of a robot head for active computer vision tasks is described. The stereo head/eye platform uses a common elevation configuration and has four degree-of-freedom. The joints are driven by DC servo motors coupled with incremental optical encoders and backlash minimizing gearboxes. The details of mechanical design, head controller design, the architecture of the system, and the design criteria for various specifications are presented.

The authors build upon a technical report (Tech. Report OUEL 2009/93, Engng. Sci., Oxford Uni., Oxford, UK, 1993) in which they proposed a model of the mammographic imaging process for which scattered radiation is a key degrading factor. Here, the authors propose a way of estimating the scatter component of the signal at any pixel within a mammographic image, and they use this estimate for model-based image enhancement. The first step is to extend the authors' previous model to divide breast tissue into "interesting" (fibrous/glandular/cancerous) tissue and fat. The scatter model is then based on the idea that the amount of scattered radiation reaching a point is related to the energy imparted to the surrounding neighbourhood. This complex relationship is approximated using published empirical data, and it varies with the size of the breast being imaged. The approximation is further complicated by needing to take account of extra-focal radiation and breast edge effects. The approximation takes the form of a weighting mask which is convolved with the total signal (primary and scatter) to give a value which is input to a "scatter function", approximated using three reference cases, and which returns a scatter estimate. Given a scatter estimate, the more important primary component can be calculated and used to create an image recognizable by a radiologist. The images resulting from this process are clearly enhanced, and model verification tests based on an estimate of the thickness of interesting tissue present proved to be very successful. A good scatter model opens the was for further processing to remove the effects of other degrading factors, such as beam hardening.

We describe the development of a monitoring system which uses sensor observation data about discrete events to construct dynamically a probabilistic model of the world. This model is a Bayesian network incorporating temporal aspects, which we call a dynamic belief network; it is used to reason under uncertainty about both the causes and consequences of the events being monitored. The basic dynamic construction of the network is data-driven. However the model construction process combines sensor data about events with externally provided information about agents' behavior, and knowledge already contained within the model, to control the size and complexity of the network. This means that both the network structure within a time interval, and the amount of history and detail maintained, can vary over time. We illustrate the system with the example domain of monitoring robot vehicles and people in a restricted dynamic environment using light-beam sensor data. In addition to presenting a generic network structure for monitoring domains, we describe the use of more complex network structures which address two specific monitoring problems, sensor validation and the data association problem.>

Abstract We demonstrate that the structure of a 3D point set with a single bilateral symmetry can be reconstructed from an uncalibrated affine image, modulo a Euclidean transformation, up to a four parameter family of symmetric objects that could have given rise to the image. If the object has two orthogonal bilateral symmetries, its shape can be reconstructed, modulo a Euclidean transformation, to a three parameter family of symmetric shapes that could have given rise to the image. Furthermore, if the camera aspects ratio is known, the three parameter family reduces to a single scale and the orientation of the object can be determined. These results are demonstrated using real images with uncalibrated cameras.

A new technique for the calibration of the intrinsic parameters of cameras for active vision systems is presented. By making deliberate camera motions, the intrinsics of the cameras can be calibrated based either on the positional difference of optical flow field (PDOFF) or the trajectories of features (TOF) on the image plane. A way to detect the distortion of a camera lens is also presented. The calibration method is simple, fast, reliable, and very easy to combine with task performing processes. The performance of the technique is illustrated. The method can be used for general calibration of camera intrinsics.>

The authors present techniques for recognizing instances of 3-D object classes from sets of 3-D feature observations. Recognition of a class instance is structured as a search of an interpretation tree in which geometric constraints on pairs of sensed features not only prune the tree, but are used to determine upper and lower bounds on the model parameter values of the instance. A real-valued constraint propagation network unifies the representations of the model parameters, model constraints and feature constraints, and provides a simple and effective mechanism for accessing and updating parameter values. Recognition of objects with multiple internal degrees of freedom, including non-uniform scaling and stretching, articulations, and subpart repetitions, is demonstrated for two different types of real range data: 3-D edge fragments from a stereo vision system, and position/surface normal data derived from planar patches extracted from a range image.>

Based on a dynamic model of a wheeled mobile robot (WMR), a dynamic control method is developed so that the WMR tracks a desired trajectory. The full nonlinear dynamics and the kinematic constraints are embodied in the controller. The steering and driving torques are chosen as the control inputs. Examples are presented that demonstrate the effectiveness of the model and the control methods.

Investigates how a car-like mobile robot handles unexpected static obstacles while following an optimal path planned by the global path planner. To find an optimal solution of the problem, the obstacle avoidance problem is formulated as a decision theoretic approach. The optimal decision rule we seek is to minimize the Bayes risk by trading off between deliberative maneuver and the alternatives. Real-time implementation is emphasized in order to provide a framework for real-world applications.

Describes a decision theoretic approach to real-time obstacle avoidance and path planning for a mobile robot. The mobile robot navigates in a semi-structured environment in which unexpected obstacles may appear at random locations. Twelve sonar sensors are currently used to report the presence and location of the obstacles. To handle the uncertainty of an obstacle's appearance, the authors adopt a Bayesian approach by assuming a prior distribution for the presence of unknown obstacles. The distribution is changed dynamically according to the information accumulated by sensors. When searching for an optimal path using dynamic programming, the authors take the probability into account in making a decision. Based on prior information and sensor data, they show that the proposed method allows the mobile robot to avoid unexpected obstacles and finds an optimal path to the goal in real time.>

This paper describes a scheme in which optical flow information is used to guide stereo correspondence matching. The epipolar constraint, which is used in static stereo to confine correspondence search to a line, is replaced by a mechanism which confines search to an area around a point. The experimental setup consists of a pair of cameras mounted on a mobile robot vehicle. A sequence of stereo image pairs is taken during vehicle motion, and the images are processed by a corner detector. One of the first steps in extracting useful information from the corners is to solve the correspondence problem. For a stereo sequence, this requires both the temporal matching of corners along the sequence, and also the stereo matching of corners between left and right images. The results of

Abstract Progress on a system to monitor the development of glaucoma by measuring the topography of the optic disk is reported. The need for an accurate method for doing this using passive vision is explained. Sparse depth measurements from stereo matching of blood vessels provide insufficient constraint for reconstructing the surface of the optic disk. Shape from shading has to contend with a spatially-varying albedo. We show how a particular arrangement of fundus cameras allows us to apply a technique akin to homomorphic filtering to recover estimates of δz δx that can be smoothed by appropriate regularization. Stereo is integrated with these photometric stereo depth estimates. Examples of artificial object and optic disk surface reconstructions are presented.

In this paper we introduce a novel representation of the significant changes in curvature along the bounding contour of planar shape. We call the representation the Curvature Primal Sketch because of the close analogy to the primal sketch representation advocated by Marr for describing significant intensity changes. We define a set of primitive parameterized curvature discontinuities, and derive expressions for their convolutions with the first and second derivatives of a Gaussian. We describe an implemented algorithm that computes the Curvature Primal Sketch by matching the multiscale convolutions of a shape, and illustrate its performance on a set of tool shapes. Several applications of the representation are sketched.

This paper reports progress toward the development of a representation of significant surface changes in dense depth maps. We call tile representation the Surface Primal Sketch by analogy with representations of intensity changes, image structure, and changes in curvature of planar curves. We describe an implemented program that detects, localizes, and symbolically describes: steps, where the surface height function is discontinuous, and roofs, where the surface is continuous but the surface normal is discontinuous. We illustrate the performance of the program on range maps of objects of varying complexity.

We summarize recent work by the author and his colleagues aimed at generating rich representations of shape for two and three-dimensional objects. An implementation of the smoothed local symmetries representation of 2D shapes is described and its use for inspection and the recognition of occluded objects sketched. The implementation uses an edge finder developed by Canny that is optimal in a precisely defined sense. The curvature patch representation of 3D shape is based on ideas of differential geometry.

An extremum principle is developed that determines three-dimensional surface orientation from a two-dimensional contour. The principle maximizes the ratio of the area to the square of the perimeter, a measure of the compactness or symmetry of the three-dimensional surface. The principle interprets regular figures correctly and it interprets skew symmetries as oriented real symmetries. The maximum likelihood method approximates the principle on irregular figures, but we show that it consistently overestimates the slant of an ellipse.

In 1963 John McCarthy proposed a formalism based on conditional expression and recursion for use in the emergent theory of computation. Included in his proposals was a proof technique, known as recursive induction, which could be used to establish the equivalence of recursively defined functions. This paper shows that the discovery of an equations to serve in a proof by recursive induction does not have to rely on luck or inspiration, but can be developed rationally hand in hand with the development of the proof.

The benefits of a programming approach to the theory of computation are illustrated by considering a traditional proof, namely the equivalence of Turing's formalism and general recursive functions (GRF). In these terms, we are led to criticise the GRF and to give a programmer's view of the Kleene normal form theorem.