Search

This paper presents a distributed and heuristic policy-based system for resource management in large-scale Grids. This approach involves three phases: resource discovery, scheduling and allocation. The resource discovery phase is supported by the SNMP-based Balanced Load Monitoring Agents for Resource Scheduling (SBLOMARS). In this approach, network and computational resources are monitored by autonomous monitoring agents, offering a pure decentralized monitoring system. The resource scheduling phase is supported by the Balanced Load Multi-Constrained Resource Scheduler (BLOMERS). It is a heuristic resource scheduler, which includes an implementation of a Genetic Algorithm (GA), as an alternative to solve the inherent NP-hard problem for resource scheduling in large-scale Grids. Allocation phase is supported by means of a Policy-based Grid Management Architecture (PbGMA). This architecture integrates different sources of service necessities such as requirements demanded by customers, applications requirements and network conditions. It interfaces with Globus middleware to allocate services into the selected resources with certain levels of QoS. [ABSTRACT FROM AUTHOR]

Next generation networks require information and communications systems able to support pervasive services and especially context-aware applications. This paper presents research challenges in self-management, autonomic communications and integration requirements of context information for supporting management operations of such services in next generation networks. The research focuses on a framework of information systems and their interoperability. Management techniques using information and data models for context information are discussed and studied and then the novel system architecture for context handling and delivery using ontology-based models is presented. In this paper, ontology-based management and modelling techniques are used and referenced in the framework of a distributed context handling and delivery system. Following this, the representative ontology-based information management system within an application scenario is presented. The flexibility of the introduced approach allows for end-user scenarios, which are briefly described at the end of the paper. [ABSTRACT FROM AUTHOR]

A novel method for simultaneously acquiring and registering range data of a real object from different viewpoints is presented. Currently, most 3D model reconstruction techniques do not cooperate with the existing range data for shape recovery of future viewpoints. In this work, a stereo vision system is developed for 3D model acquisition. To reduce the computation and increase the accuracy of stereo matching algorithms, the recovered range data from previous viewpoints are registered and then used to provide additional constraints for 3D acquisition of the next viewpoint. Experiments have shown that our approach gives better performance on both execution time and stereo matching results. [ABSTRACT FROM AUTHOR]

This paper reports the results obtained by analysing some of the most well-known features used in Computer Vision to describe and classify shapes in an appealing real application. We aim to demonstrate the applicability of shape descriptors to classify muscles on Magnetic Resonance Imaging (MRI). The mechanized classification of ham muscles could help the industries to automate the ripening process for Iberian ham. The excellent classification percentages obtained in our experiments suggest the real viability of the feature vector developed in this paper to recognize and classify muscles. [ABSTRACT FROM AUTHOR]

We address the problem of estimating 3-D motion from acoustic images acquired by high-frequency 2-D imaging sonars deployed in underwater. Utilizing a planar approximation to scene surfaces, two-view homography is the basis of a nonlinear optimization method for estimating the motion parameters. There is no scale factor ambiguity, unlike the case of monocular motion vision for optical images. Experiments with real images demonstrate the potential in a range of applications, including target-based positioning in search and inspection operations. [ABSTRACT FROM AUTHOR]

Coarseness is a very important textural concept that has been widely analyzed in computer vision for years. However, a model which allows to represent different perception degrees of this textural concept in the same way that humans perceive texture is needed. In this paper we propose a model that associates computational measures to human perception by learning an appropriate function. To do it, different measures representative of coarseness are chosen and subjects assessments are collected and aggregated. Finally, a function that relates these data is fitted. [ABSTRACT FROM AUTHOR]

In this article, scale and orientation invariant object detection is performed by matching intensity level histograms. Unlike other global measurement methods, the present one uses a local feature description that allows small changes in the histogram signature, giving robustness to partial occlusions. Local features over the object histogram are extracted during a Boosting learning phase, selecting the most discriminant features within a training histogram image set. The Integral Histogram has been used to compute local histograms in constant time. [ABSTRACT FROM AUTHOR]

This paper extends and generalizes the Bayesian semi-supervised segmentation algorithm [1] for oil spill detection using SAR images. In the base algorithm on which we build on, the data term is modeled by a finite mixture of Gamma distributions. The prior is an M-level logistic Markov Random Field enforcing local continuity in a statistical sense. The methodology proposed in [1] assumes two classes and known smoothness parameter. The present work removes these restrictions. The smoothness parameter controlling the degree of homogeneity imposed on the scene is automatically estimated and the number of used classes is optional. Semi-automatic estimation of the class parameters is also implemented. The maximum a posteriori (MAP) segmentation is efficiently computed via the α− expansion algorithm [2], a recent graph-cut technique, The effectiveness of the proposed approach is illustrated with simulated (Gaussian or Gamma data term and M-level logistic classes) and real ERS data. [ABSTRACT FROM AUTHOR]

Tracking the Left Ventricle (LV) in ultrasound sequences remains a challenge due to speckle noise, low SNR and lack of contrast. Therefore, it is usually difficult to obtain accurate estimates of the LV cavities since feature detectors produce a large number of outliers. This paper presents an algorithm which combines two main operations: i) a novel denoising algorithm based on the Lyapounov equation and ii) a robust tracker, based on an outlier feature model. Experimental results are provided, showing that the proposed algorithm is computationally efficient and leads to accurate estimates of the LV. [ABSTRACT FROM AUTHOR]

We present a new method for the synchronization of a pair of video sequences and the spatial registration of all the temporally corresponding frames. This is a mandatory step to perform a pixel wise comparison of a pair of videos. Several proposals for video matching can be found in the literature, with a variety of applications like object detection, visual sensor fusion, high dynamic range and action recognition. The main contribution of our method is that it is free from three common restrictions assumed in previous works. First, it does not impose any condition on the relative position of the two cameras, since they can move freely. Second, it does not assume a parametric temporal mapping relating the time stamps of the two videos, like a constant or linear time shift. Third, it does not rely on the complete trajectories of image features (points or lines) along time, something difficult to obtain automatically in general. We present our results in the context of the comparison of videos captured from a camera mounted on moving vehicles. [ABSTRACT FROM AUTHOR]

Hidden Markov Models (HMM) have been successfully applied to describe sequences of observable events. In some problems, objects are more appropriately described as cyclic sequences, i.e., sequences with no begin/end point. Conventional HMMs with Viterbi score cannot deal adequately with cyclic sequences. We propose a cyclic Viterbi score that can be efficiently computed for Linear HMMs. Linear HMMs model sequences that can be partitioned into contiguous segments where each state is responsible for emitting all symbols in one of the segments. Experiments show that our proposal outperforms other approaches in an isolated characters handwritten-text recognition task. [ABSTRACT FROM AUTHOR]

Linear unmixing decomposes a hyperspectral image into a collection of reflectance spectra of the materials present in the scene, called endmember signatures, and the corresponding abundance fractions at each pixel in a spatial area of interest. This paper introduces a new unmixing method, called Dependent Component Analysis (DECA), which overcomes the limitations of unmixing methods based on Independent Component Analysis (ICA) and on geometrical properties of hyperspectral data. DECA models the abundance fractions as mixtures of Dirichlet densities, thus enforcing the constraints on abundance fractions imposed by the acquisition process, namely non-negativity and constant sum. The mixing matrix is inferred by a generalized expectation-maximization (GEM) type algorithm. The performance of the method is illustrated using simulated and real data. [ABSTRACT FROM AUTHOR]

This paper describes an approach to estimate the parameters of a motion blur (direction and length) directly form the observed image. The motion blur estimate can then be used in a standard non-blind deconvolution algorithm, thus yielding a blind motion deblurring scheme. The estimation criterion is based on recent results about the general spectral behavior of natural images. Experimental results show that the proposed approach is able to accurately estimate both the length and orientation of motion blur kernels, even for small lengths which are traditionally difficult. [ABSTRACT FROM AUTHOR]

Shape elongation is one of the basic shape descriptors that has a very clear intuitive meaning. That is reason for its applicability in many shape classification tasks. In this paper we define a new method for computing shape elongation for shapes with polygonal boundaries. The measure is the ratio of the maximal and minimal of the sums of squared lengths of the projections of all of the edges of the polygonal boundary onto a line which has a particular slope. We express the measure with a closed formula. This measure finds the elongation for shapes whose boundary is not extracted completely, which is impossible to achieve with existing area based measures. [ABSTRACT FROM AUTHOR]

The median graph is a useful tool to cluster a set of graphs and obtain a prototype of them. The spectral graph theory is another approach to represent graphs and find "good" approximate solutions for the graph-matching problem. Recently, both approaches have been put together and a new representation has emerged, which is called Spectral-Median Graphs. In this paper, we summarize and compare two techniques to synthesize a Spectral-Median Graph: one is based on the correlation of the modal matrices and the other one is based on the averaging of the spectral modes. Results show that, although both approaches obtain good prototypes of the clusters, the first one is slightly more robust against the noise than the second one. [ABSTRACT FROM AUTHOR]

This paper presents a novel methodology to obtain the breast skin line in mammographic images. The breast edge provides important information of the breast shape and deformation which is posteriorly used by other processing techniques, typically mammographic image registration and abnormality detection. The proposed methodology is based on applying edge detection algorithms and scale space concepts. The proposed method is a particular implementation (application focused) of a growing active contour with common considerations. Quantitative and qualitative evaluation is provided to show the validity of the approach. [ABSTRACT FROM AUTHOR]

Automatic Speech/Music Discrimination (SMD) has become a research topic of interest in the last years. This paper present a new approach for such goal, which is mainly based on a distributed expert system that incorporates fuzzy rules into its knowledge base. The proposed SMD scheme consists of two stages: 1) features extraction, 2) classification of parameters. Classification is performed by cascading a GMM-based classifier with an Evolutionary Fuzzy Expert (EFE) system. The EFE system improves the accuracy rate provided by the GMM-based classifier taking into account information of current and past audio frames. Testing the kindness of new fuzzy rules for the expert system has a high computacional cost. For that reason, a distributed learning approach based on web services has been implemented. [ABSTRACT FROM AUTHOR]

In traffic sign recognition systems, one of the normal approaches is the identification of the shape of the sign prior to the recognition itself. Normally, the recognition process needs an accurate localization of the sign for a good performance. If we are dealing with triangular, rectangular and circular signs, this means the accurate localization of the vertices of the triangle and the rectangle, or the parameters of the ellipse. In this paper we have developed a system which searches the above mentioned parameters from the signature of the blob using techniques of algebraic-distance minimization. Comparisons with previous works show good improvements in the localization of the shape, especially in the presence of slight occlusions. This work is part of a traffic sign recognition system, and in this paper we focus on the shape localization step. [ABSTRACT FROM AUTHOR]

Colposcopy test is the second most used technique to diagnose cervical cancer disease. Some researchers have proposed to use temporal changes intrinsic to the colposcopic image sequences to automatically characterize cervical lesion. Under this approach, every single pixel on the image is represented as a Time Series of length equal to the sampling frequency times acquisition points. Although this approach seems to show promising results, the data analysis procedures have to deal with huge data set that rapidly increase with the number of cases (patients) considered in the analysis. In the present work, we perform principal component analysis (PCA) to reduce the dimensionality of the data in order to facilitate similarity measures for classification and clustering. The importance of this work is that we propose a model to parameterize the dynamics of the system using an efficient representation getting a 1.11% data compression ratio and similarity on clustering of 0.78. The feasibility of the proposed model is shown testing the similarity of the clusters generated using the k-means algorithm over the raw data and the compressed representation of real data. [ABSTRACT FROM AUTHOR]

Case-Based Reasoning (CBR) systems solve new problems using others which have been previously resolved. The knowledge is composed of a set of cases stored in a case memory, where each one describes a situation in terms of a set of features. Therefore, the size and organization of the case memory influences in the computational time needed to solve new situations. We organize the memory using Self-Organization Maps, which group cases with similar properties into patterns. Thus, CBR is able to do a selective retrieval using only the cases from the most suitable pattern. However, the data complexity may hinder the identification of patterns and it may degrade the accuracy rate. This work analyses the successful application of this approach by doing a previous data complexity characterization. Relationships between the performance and some measures of class separability and the discriminative power of attributes are also found. [ABSTRACT FROM AUTHOR]

Carotid atherosclerosis is the most common life-threatening neurological disease and therefore an accurate assessment of atheromatous plaques is clinically important. Several studies were developed to characterize plaques from two-dimensional (2D) ultrasound images that are associated with high risk of stroke. However, 2D characterization may not be very accurate because it depends on the selection of a representative ultrasound image of the plaque by an experimented physician. In this paper we present a novel approach for diagnosis based on 3D ultrasound, which only requires a common ultrasound equipment without need of any additional and expensive devices like spatial locators. The semi-automatic algorithm uses medical guidance to obtain a 3D representation of the carotid artery and plaque and automatically generates measures to characterize the plaque in terms of dimensions and texture. A useful analysis tool is provided to allow the identification of vulnerable foci within the plaque. [ABSTRACT FROM AUTHOR]

This paper describes a new aerial images segmentation algorithm. Kernel Matching Pursuit (KMP) method is introduced to deal with the nonlinear distribution of the man-made objects' features in the aerial images. In KMP algorithm, a lot of training samples containing substantive information are used to detect the man-made objects. With KMP classifier, pixels in large aerial images will be labeled as different prediction values, which can be classified linearly. Then the modified Mumford-Shah model, which comprises the features of the KMP prediction values, is built to segment the aerial image by necessary level set evolution. The proposed method is proven to be effective by the results of experiments. [ABSTRACT FROM AUTHOR]

We describe a new approach to incorporate adaptivity into the partial differential equations (PDEs) of morphological dilation and erosion. By multiplication of the image gradient with a space-variant matrix, the speed of the evolution is locally adapted to the data. This is used to create anisotropic morphological evolutions that enhance coherent, flow-like image structures. We show that our adaptive method can be implemented by means of a simple modification of the classical Rouy-Tourin finite difference scheme. Numerical experiments confirm that the proposed dilations and erosions are capable of real anisotropic behaviour that can be used for closing interrupted lines. [ABSTRACT FROM AUTHOR]

This paper makes use of several performance metrics to extend the understanding of a challenging imbalanced classification task. More specifically, we refer to a problem in which the minority class is more represented in the overlap region than the majority class, that is, the overall minority class becomes the majority one in this region. The experimental results demonstrate that the use of a set of appropriate performance measures allows to figure out such an atypical case. [ABSTRACT FROM AUTHOR]

Median graphs have been presented as an useful tool for capturing the essential information of a set of graphs. The computation of the median graph is a complex task. Exact algorithms are, in the worst case, exponential both in the number of graphs and their size. The known bounds for the minimum and maximum number of nodes of the candidate median graphs are in general very coarse and they can be used to achieve only limited improvements in such algorithms. In this paper we present more accurate bounds based on the well-known concepts of maximum common subgraph and minimum common supergraph. These new bounds on the number of nodes can be used to improve the existing algorithms in the computation of the median graph. [ABSTRACT FROM AUTHOR]

Random forest is a collection (ensemble) of decision trees. It is a popular ensemble technique in pattern recognition. In this article, we apply random forest for cancer classification based on gene expression and address two issues that have been so far overlooked in other works. First, we demonstrate on two different real-world datasets that the performance of random forest is strongly influenced by dataset complexity. When estimated before running random forest, this complexity can serve as a useful performance indicator and it can explain a difference in performance on different datasets. Second, we show that one should rely with caution on feature importance used to rank genes: two forests, generated with the different number of features per node split, may have very similar classification errors on the same dataset, but the respective lists of genes ranked according to feature importance can be weakly correlated. [ABSTRACT FROM AUTHOR]

The basis for the high-level interpretation of observed patterns of human motion still relies on motion segmentation. Popular approaches based on background subtraction use colour information to model each pixel during a training period. Nevertheless, a deep analysis on colour segmentation problems demonstrates that colour segmentation is not enough to detect all foreground objects in the image, for instance when there is a lack of colour necessary to build the background model. In this paper, our segmentation procedure is based not only on colour, but also on intensity information. Consequently, the intensity model enhances segmentation when the use of colour is not feasible. Experimental results demonstrate the feasibility of our approach. [ABSTRACT FROM AUTHOR]

In this paper, we address the topic of how to estimate phrase-based models from very large corpora and apply them in statistical machine translation. The great number of sentence pairs contained in recent corpora like the well-known Europarl corpus have enormously increased the memory requirements to train phrase-based models and to apply them within a decoding process. We propose a general framework that deals with this problem without introducing significant time overhead by means of the combination of different scaling techniques. This new framework is based on the use of counts instead of probabilities, and on the concept of cache memory. [ABSTRACT FROM AUTHOR]

Mathematical Morphology is a powerful non-linear image analysis techniques based on lattice theory. The definitions of morphological operators need an ordered lattice algebraic structure. In order to apply these operators to the colour images it is required, on one hand the choice of a suitable colour space representation and on the other hand, to establish an order in the colour space providing an ordered lattice algebraic structure. The HSI space represents the colour in terms of physical attributes that separate the achromatic component from the chromatic one and it yields a more intuitive description of the colour properties than the RGB space. The suggested order weighs the hue and the intensity according to the saturation level: it has a lexicographical order in which the intensity has priority if the saturation is high, and the hue has priority if the saturation is low. [ABSTRACT FROM AUTHOR]

This paper describes how face recognition can be effected using 3D shape information extracted from single 2D image views. We characterise the shape of the field of facial normals using a statistical model based on principal geodesic analysis. The model can be fitted to 2D brightness images of faces to recover a vector of shape parameters. Since it captures variations in a field of surface normals, the dimensionality of the shape vector is twice the number of image pixels. We investigate how to perform face recognition using the output of PGA by applying a number of dimensionality reduction techniques including principal components analysis, locally linear embedding, locality preserving projection and Isomap. [ABSTRACT FROM AUTHOR]

This work focuses on the development of a computer vision system for the automatic on-line inspection and classification of Satsuma segments. During the image acquisition the segments are in movement, wet and frequently in contact with other pieces. The segments are transported over six semi-transparent conveyor belts that advance at speed of 1 m/s. During on-line operation, the system acquires images of the segments using two cameras connected to a single computer and process the images in less than 50 ms. Extracting morphological features from the objects, the system identifies automatically pieces of skin and row material and separates entire segments from broken ones, discriminating between those with slight or large breaking degree. Combinations of morphological parameters were employed to decide the quality of each segment, classifying correctly 95% of sound segments. [ABSTRACT FROM AUTHOR]

We propose a data-driven approach for the Jacobian adaptation (JA) to make it more robust against the noisy environments in speech recognition. The reference hidden Markov model (HMM) in the JA is trained directly with the noisy speech for improved acoustic modeling instead of using the model composition methods like the parallel model combination (PMC). This is made possible by estimating the Jacobian matrices and other statistical information for the adaptation using the Baum-Welch algorithm during the training. The adaptation algorithm has shown to give improved robustness especially when used in a multi-model structure. From the speech recognition experiments based on HMMs, we could find the proposed adaptation method gives better recognition results compared with conventional HMM parameter compensation methods and the multi-model approach could be a viable solution in the noisy speech recognition. [ABSTRACT FROM AUTHOR]

In this paper we address the problem of on-line recognition of human activities taking place in a public area such as a shopping center. We consider standard activities; namely, entering, exiting, passing or browsing. The problem is motivated by surveillance applications, for which large numbers of cameras have been deployed in recent years. Such systems should be able to detect and recognize human activities, with as little human intervention as possible. In this work, we model the displacement of a person in consecutive frames using a bank of switched dynamical systems, each of which tailored to the specific motion regimes that each trajectory may contain. Our experimental results are based on nearly 20,000 images concerning four atomic activities and several complex ones, and demonstrate the effectiveness of the proposed approach. [ABSTRACT FROM AUTHOR]

Three methods for the efficient downdating, composition and splitting of low rank singular value decompositions are proposed. They are formulated in a closed form, considering the mean information and providing exact results. Although these methods are presented in the context of computer vision, they can be used in any field forgetting information, combining different eigenspaces in one or ignoring particular dimensions of the column space of the data. Application examples on face subspace learning and latent semantic analysis are given and performance results are provided. [ABSTRACT FROM AUTHOR]

Minimally Invasive Surgery (MIS) is one of these applications where usually only 2D information is available to perform a 3D task. It requires a high degree of sensory-motor skills to overcome the disengagement between action and perception caused by the physical separation of the surgeon with the operative site. The integration of body movements with visual information serves to assist the surgeon providing a sense of position. Our purpose in this paper is to present a solution to the exterior orientation problem based on computer vision, as a tool in assisted interventions, locating the instruments with respect to the surgeon. Having knowledge of the 3D transformations applied to the instrument and its projections in the image plane, we show it is possible to estimate its orientation with only two different rotations and also its relative position if scale information is supplied. Experimental results show some advantages of this new algorithm such as simplicity and real-time performance. [ABSTRACT FROM AUTHOR]