Search

Abstract: This paper proposes two efficient kernels for comparing acyclic, directed graphs. The first kernel counts the number of common paths and allows for weighing according to path-length and/or according to the vertices contained in each particular path. The second kernel counts the number of paths in common minors of the graphs involved and allows for length- and vertex-weighting too. Both kernels have algorithmic complexity that is cubic in the size of the vertex-set. The performance of the algorithms is concisely demonstrated using synthetic and real data. [Copyright &y& Elsevier]

Abstract: Point cloud matching is a central problem in Object Modeling with applications in Computer Vision and Computer Graphics. Although the problem is well studied in the case when an initial estimate of the relative pose is known (fine matching), the problem becomes much more difficult when this a priori knowledge is not available (coarse matching). In this paper we introduce a novel technique to speed up coarse matching algorithms for point clouds. This new technique, called Hierarchical Normal Space Sampling (HNSS), extends Normal Space Sampling by grouping points hierarchically according to the distribution of their normal vectors. This hierarchy guides the search for corresponding points while staying free of user intervention. This permits to navigate through the huge search space taking advantage of geometric information and to stop when a sufficiently good initial pose is found. This initial pose can then be used as the starting point for any fine matching algorithm. Hierarchical Normal Space Sampling is adaptable to different searching strategies and shape descriptors. To illustrate HNSS, we present experiments using both synthetic and real data that show the computational complexity of the problem, the computation time reduction obtained by HNSS and the application potentials in combination with ICP. [Copyright &y& Elsevier]

Abstract: Causal structure learning algorithms construct Bayesian networks from observational data. Using non-interventional data, existing constraint-based algorithms may return I-equivalent partially directed acyclic graphs. However, these algorithms do not fully exploit the graphical properties of Bayesian networks, and require many redundant tests that reduce both speed and accuracy. In this paper, we introduce ideas to exploit such properties to increase the speed and accuracy of causal structure learning for multivariate normal data. In numerical experiments on five benchmarking networks our proposed algorithm was faster and more accurate than recently-developed algorithms. [Copyright &y& Elsevier]

Abstract: Object discovery in visual object categorisation (VOC) is the problem of automatically assigning class labels to objects appearing in given images. To achieve state-of-the-art results in this task, a large set of positive and negative training images from publicly available benchmark data sets have been used to train discriminative classification methods. The immediate drawback of these methods is the requirement of a vast amount of labelled data. Therefore, the ultimate challenge for visual object categorisation has been recently exposed: unsupervised object discovery, also called unsupervised VOC (UVOC), where the selection of the number of classes and the assignments of given images to these classes are performed automatically. The problem is very challenging and hitherto only a few methods have been proposed. These methods are based on the popular bag-of-features approach and clustering to automatically form the classes. In this paper, we adopt the self-organising principle and replace clustering with the self-organising map (SOM) algorithm. Our method provides results comparable to the state of the art and its advantages, such as non-sensitivity against codebook histogram normalisation, advocate its usage in unsupervised object discovery. [Copyright &y& Elsevier]

Abstract: Curvilinear structures are useful features in a variety of applications, particularly in medical image analysis. Compared to other commonly used features such as edges and regions, there is relatively few work on performance evaluation methodologies for curvilinear structure detection algorithms. For instance, a pixel-wise comparison with ground truth has been used in all recent publications on vessel detection in retinal images. In this paper we propose a novel structure-based methodology for evaluating the performance of 2D and 3D curvilinear structure detection algorithms. We consider the two aspects of performance, namely detection rate and detection accuracy, separately, in contrast to their mixed handling in earlier approaches that typically produces biased impression of detection quality. By doing so, the proposed performance measures give us a more informative and precise performance characterization. Experiments on both synthetic and real examples will be given to demonstrate the advantages of our approach. [Copyright &y& Elsevier]

Abstract: Graph comparison algorithms based on metric space embedding have been proven robust in graph clustering and classification. In this paper we propose graph embedding method exploiting ordered invariants of distance k-graphs, which encode structure of shortest-paths. We study degree histograms of those graphs and use them to construct permutation invariant graph representation called vertex B-matrix. In order to extract more information from structural patterns we also define edge distance k-graphs and associated edge B-matrix. Next, several new graph characteristics obtained by condensing information stored in B-matrices are introduced. We demonstrate that our approach provides stable embedding, which captures relevant graph features. Experiments on classification with satellite photo and mutagenicity benchmark datasets revealed, that new descriptors allow for distinguishing graphs with non trivial structural differences. Moreover, they appear to outperform descriptors based on heat kernel matrix, being at the same time more effective computationally. In the end we test feature selection on B-matrices showing that selecting right B-submatrix can improve classification rate on testing datasets. [Copyright &y& Elsevier]

Abstract: In this paper, we present a hypergraph based geometric biclustering (HGBC) algorithm. In a high dimensional space, bicluster patterns to be recognized can be considered to be linear geometrical structures. We can use the Hough transform (HT) to find sub-biclusters which correspond to the linear structures in column-pair spaces. Then a hypergraph model is built to merge the sub-biclusters into larger ones. Experiments on simulated and real biological data show that the HGBC algorithm proposed here can combine the sub-biclusters efficiently and provide more accurate classification results compared with existing biclustering methods. [Copyright &y& Elsevier]

Abstract: Morse theory is a fundamental tool for analyzing the geometry and topology of smooth manifolds. This tool was translated by Forman to discrete structures such as cell complexes, by using discrete Morse functions or equivalently gradient vector fields. Once a discrete gradient vector field has been defined on a finite cell complex, information about its homology can be directly deduced from it. In this paper we introduce the foundations of a homology-based heuristic for finding optimal discrete gradient vector fields on a general finite cell complex K. The method is based on a computational homological algebra representation (called homological spanning forest or HSF, for short) that is an useful framework to design fast and efficient algorithms for computing advanced algebraic-topological information (classification of cycles, cohomology algebra, homology A(∞)-coalgebra, cohomology operations, homotopy groups, …). Our approach is to consider the optimality problem as a homology computation process for a chain complex endowed with an extra chain homotopy operator. [Copyright &y& Elsevier]

Abstract: In this paper we present a novel algorithm to optimize the reconstruction from non-uniform point sets. We introduce a statistically-derived topology-controller for selecting the reconstruction resolution of a given non-uniform point set. Deriving information from homology-based statistics, our topology-controller ensures a stable and sound basis for the analysis process. By analyzing our topology-controller, we select an optimal reconstruction resolution which ensures both low reconstruction errors and a topological stability of the underlying signal. Our approach offers a valuable method for the evaluation of the reconstruction process without the need of visual inspection of the reconstructed datasets. By means of qualitative results we show how our proposed topology statistics provides complementary information in the enhancement of existing reconstruction pipelines in visualization. [Copyright &y& Elsevier]

Abstract: Compressive sensing is an emerging field predicated upon the fact that, if a signal has a sparse representation in some basis, then it can be almost exactly reconstructed from very few random measurements. Many signals and natural images, for example under the wavelet basis, have very sparse representations, thus those signals and images can be recovered from a small amount of measurements with very high accuracy. This paper is concerned with the dimensionality reduction problem based on the compressive assumptions. We propose novel unsupervised and semi-supervised dimensionality reduction algorithms by exploiting sparse data representations. The experiments show that the proposed approaches outperform state-of-the-art dimensionality reduction methods. [Copyright &y& Elsevier]

Abstract: Two novel methods for extracting distinctive invariant features from interest regions are presented in this paper. The idea of these methods are associated with that measuring similarity between visual entities from images can be based on matching the internal layout of Local Self-Similarities. The main contributions are two-folds: firstly, two new texture features called Local Self-Similarities (LSS,C) and Fast Local Self-Similarities (FLSS,C) based on Cartesian location grid, are extracted, which are the modified versions of the well-known Local Self-Similarities (LSS,LP) feature based on Log-Polar location grid. To combine the powers of the SIFT and LSS (LP), LSS and FLSS are used as the local features in the SIFT algorithm. Secondly, different from the natural LSS (LP) descriptor that chooses the maximal correlation value in each bucket to get photometric translations invariance, the proposed LSS (C) and FLSS (C) adopt distribution-based representation to achieve more robust geometric translations invariance. In the contexts of image matching and object category classification experiments, the LSS (C) and FLSS (C) both outperform the original LSS (LP), and achieve favorably comparable performance to the SIFT. Furthermore, these descriptors are low computational complexity and simpler than the SIFT. [Copyright &y& Elsevier]

Abstract: This paper examines a problem in the multi-model representation of digital curves. It presents Dynamic Programming algorithms for curves approximation with a Minimum Description Length for a given error threshold with measure L ∞ or L 2. For the error measure L ∞, the optimal algorithm was based on a search for the shortest path in the weighted multigraph constructed on the vertices of the curve. As for the case with an approximation with L 2-norm, the optimal algorithm includes the construction of the shortest path in two-dimensional search space. We then proposed various fast and efficient versions of the algorithms for the solution of the problem. We proceeded to test these algorithms on large-size contours and were able to demonstrate a good trade-off between time performance and the efficiency of the solutions. We were thus able to produce results for the optimal and fast near-optimal algorithms for a two-model approximation with line segments and circular arcs. In addition, the proposed algorithm was demonstrated on the adaptive motion model for trajectory segmentation. [Copyright &y& Elsevier]

Abstract: It has been showed that, by employing fake fingers, the existing fingerprint recognition systems may be easily deceived. So, there is an urgent need for improving their security. Software-based liveness detection algorithms typically exploit morphological and perspiration-based characteristics separately to measure the vitality. Both such features provide discriminant information about live and fake fingers, then, it is reasonable to investigate also their joint contribution. In this paper, we combine a set of the most robust morphological and perspiration-based measures. The effectiveness of the proposed approach has been assessed through a comparison with several state-of-the-art techniques for liveness detection. Experiments have been carried out, for the first time, by adopting standard databases. They have been taken from the Liveness Detection Competition 2009 whose data have been acquired by using three different optical sensors. Further, we have analyzed how the performance of our algorithm changes when the material employed for the spoof attack is not available during the training of the system. [Copyright &y& Elsevier]

Abstract: This paper presents a weighted co-occurrence phase histogram (WCPH) for representing the local characteristics of texture pattern and applies it to iris recognition. We first introduce a weighting function that enables the phase angle of the image gradient at one pixel to contribute smoothly to several adjacent histogram bins. This accounts for the uncertainty of the phase angle estimation brought by disturbing factors such as noise and illumination changes. The weighting function also avoids the quantization problem typical of the traditional histogram. We then define the WCPH by computing the weighted co-occurrence of pairs of image pixels that are at fixed distance. The WCPH models the joint probability distribution of both the phase angle and spatial layout, thus having the potential to capture richer information in texture pattern. Based on the WCPH, we develop an iris recognition algorithm using the Bhattacharyya distance to measure the goodness of match. The recognition algorithm considers the effects of noise and employs a simple image registration scheme to account for image deformation. We evaluate the performance of the proposed work on the UBIRIS.v2 database. We participated in the Noisy Iris Challenge Evaluation-Part II (NICE:II). It evaluates the robustness to noise of iris encoding and matching methods on the UBIRIS.v2 database, where the iris images are captured at-the-distance and on-the-move. We ranked #5 among all the registered participants according to the evaluation of the NICE:II organizing committee. [Copyright &y& Elsevier]

Abstract: This paper studies the iris recognition problem in the degraded iris images captured in non-ideal imaging conditions. In these circumstances iris recognition becomes challenging because of noisy factors such as the off-axis imaging, pose variation, image blurring, illumination change, occlusion, specular highlights and noise. We introduce a robust algorithm based on the Random Sample Consensus (RANSAC) for localization of non-circular iris boundaries. It can localize the iris boundaries more accurately than the methods based on the Hough transform. To account for iris pattern deformation, we describe an image registration method based on the Lucas–Kanade algorithm. Operating on the filtered iris images, this method divides one image into small sub-images and solves registration problem for every small sub-image. Under some reasonable assumptions this method becomes very efficient while maintaining its effectiveness. Finally, we investigate how to extract highly distinctive features in the degraded iris images. We present a sequential forward selection method for seeking a sub-optimal subset of filters from a family of Gabor filters. The recognition performance is greatly improved with a very small number of filters selected. Experiments were conducted on the UBIRIS.v2 iris database and promising results were obtained. [Copyright &y& Elsevier]

Abstract: Noisy iris recognition under visible lighting has recently drawn much attention. This paper proposes an effective method for visible light iris image matching by using multiple characteristics of iris and eye images. The method consists of image preprocessing, iris data matching, eye data matching, and multi-modal fusion. Ordinal measures and color analysis are adopted for iris data matching, and texton representation and semantic information are used for eye data matching. After we obtain the four matching scores, a robust score level fusion strategy is applied to generate the dissimilarity measure of the two images under consideration. Extensive experiments on the UBIRIS.v2 database and the NICE.II training dataset demonstrate that the proposed method is effective. Our method significantly outperforms all other algorithms submitted to the Noisy Iris Challenge Evaluation-Part II (NICE.II), an open contest in noisy iris image matching. [Copyright &y& Elsevier]

Abstract: This paper presents a novel method for addressing the problem of finding more good feature pairs between images, which is one of the most fundamental tasks in computer vision and pattern recognition. We first select matched features by Bi-matching as seed points, then organize these seed points by adopting the Delaunay triangulation algorithm. Finally, triangle constraint is used to explore good matches. The experimental evaluation shows that our method is robust to most geometric and photometric transformations including rotation, scale change, blur, viewpoint change, JPEG compression and illumination change, and significantly improves both the number of correct matches and the matching score. And the application on estimating the fundamental matrix for a pair of images is also shown. Both the experiments and the application demonstrate the robust performance of our method. [Copyright &y& Elsevier]

Abstract: The field of pattern recognition is usually subdivided into the statistical and the structural approach. Structural pattern recognition allows one to use powerful and flexible representation formalisms but offers only a limited repertoire of algorithmic tools needed to solve classification and clustering problems. By contrast, the statistical approach is mathematically well founded and offers many tools, but provides a representation formalism that is limited in its power and flexibility. Hence, both subfields are complementary to each other. During the last three decades several efforts have been made towards bridging the gap between structural and statistical pattern recognition in order to profit from the benefits of each approach and eliminate the drawbacks. The present paper reviews some of these attempts made towards the unification of structural and statistical pattern recognition and analyzes the progress that has been achieved. [Copyright &y& Elsevier]

Abstract: Cast shadows need careful consideration in the development of robust dynamic scene analysis systems. Cast shadow detection is critical for accurate object detection in video streams, and their misclassification can cause errors in segmentation and tracking. Many algorithms for shadow detection have been proposed in the literature; however a complete, comparative evaluation of existing approaches is lacking. This paper presents a comprehensive survey of shadow detection methods, organised in a novel taxonomy based on object/environment dependency and implementation domain. In addition a comparative evaluation of representative algorithms, based on quantitative and qualitative metrics is presented to evaluate the algorithms on a benchmark suite of indoor and outdoor video sequences. [Copyright &y& Elsevier]

Abstract: In this paper, we propose a wavelet-based eigentransformation method for human face hallucination. Our algorithm uses the wavelet transform to decompose interpolated low-resolution (LR) images in the wavelet domain to obtain high-frequency information in three different directions, and employs the eigentransformation method to reconstruct the corresponding finer high-frequency content of the high-resolution (HR) images. The low-frequency content of the HR images in the wavelet domain is estimated based on the interpolated images directly. The resulting high-quality HR faces can be synthesized by using the inverse wavelet transform, with all the estimated coefficients. By combining interpolation and eigentransformation, the reconstructed images are less dependent on the training set selected, and can better preserve the low-frequency content. Thus, the reconstructed images look more like the ground-true HR images, as compared to the original eigentransformation method. Experimental results show that our proposed algorithm outperforms the original eigentransformation and other existing methods for face hallucination in terms of both visual quality and objective measurements. [Copyright &y& Elsevier]

Abstract: Pattern recognition using Dynamic Bayesian Networks (DBNs) is currently a growing area of study. The classification performance greatly relies on the choice of a DBN model that will best describe the dependencies in each class of data. In this paper, we present DBN models trained for the classification of handwritten digit. Two approaches to improve the suitability of the models are presented. One uses a fixed DBN structure, and is based on an Evolutionary Algorithm optimisation of the selection and of the layout of the observations for each class of data. The second approach is about learning part of the structure of the models from the training set of each class. Parameter learning is then performed for each DBN. Classification results are presented for the described models, and compared with previously published results. Both approaches were found to improve the recognition rate compared to previous results. [Copyright &y& Elsevier]

Abstract: Both the quality and quantity of training data have significant impact on the accuracy of rank functions in web search. With the global search needs, a commercial search engine is required to expand its well tailored service to small countries as well. Due to heterogeneous intrinsic of query intents and search results on different domains (i.e., for different languages and regions), it is difficult for a generic ranking function to satisfy all type of queries. Instead, each domain should use a specific well tailored ranking function. In order to train each ranking function for each domain with a scalable strategy, it is critical to leverage existing training data to enhance the ranking functions of those domains without sufficient training data. In this paper, we present a boosting framework for learning to rank in the multi-task learning context to attack this problem. In particular, we propose to learn non-parametric common structures adaptively from multiple tasks in a stage-wise way. An algorithm is developed to iteratively discover super-features that are effective for all the tasks. The estimation of the regression function for each task is then learned as linear combination of those super-features. We evaluate the accuracy of multi-task learning methods for web search ranking using data from multiple domains from a commercial search engine. Our results demonstrate that multi-task learning methods bring significant relevance improvements over existing baseline method. [Copyright &y& Elsevier]

Abstract: This paper describes a corpus, dataset and associated ground-truth, for the evaluation of people detection algorithms in surveillance video scenarios, along with the design procedure followed to generate it. Sequences from scenes with different levels of complexity have been manually annotated. Each person present at a scene has been labeled frame by frame, in order to automatically obtain a people detection ground-truth for each sequence. Sequences have been classified into different complexity categories depending on critical factors that typically affect the behavior of detection algorithms. The resulting corpus, which exceeds other public pedestrian datasets in the amount of video sequences and its complexity variability, is freely available for benchmarking and research purposes under a license agreement. [Copyright &y& Elsevier]

Abstract: Isometric mapping (Isomap) is a popular nonlinear dimensionality reduction technique which has shown high potential in visualization and classification. However, it appears sensitive to noise or scarcity of observations. This inadequacy may hinder its application for the classification of microarray data, in which the expression levels of thousands of genes in a few normal and tumor sample tissues are measured. In this paper we propose a double-bounded tree-connected variant of Isomap, aimed at being more robust to noise and outliers when used for classification and also computationally more efficient. It differs from the original Isomap in the way the neighborhood graph is generated: in the first stage we apply a double-bounding rule that confines the search to at most k nearest neighbors contained within an ε-radius hypersphere; the resulting subgraphs are then joined by computing a minimum spanning tree among the connected components. We therefore achieve a connected graph without unnaturally inflating the values of k and ε. The computational experiences show that the new method performs significantly better in terms of accuracy with respect to Isomap, k-edge-connected Isomap and the direct application of support vector machines to data in the input space, consistently across seven microarray datasets considered in our tests. [Copyright &y& Elsevier]

Abstract: This paper presents a novel local threshold algorithm for the binarization of document images. Stroke width of handwritten and printed characters in documents is utilized as the shape feature. As a result, in addition to the intensity analysis, the proposed algorithm introduces the stroke width as shape information into local thresholding. Experimental results for both synthetic and practical document images show that the proposed local threshold algorithm is superior in terms of segmentation quality to the threshold approaches that solely use intensity information. [Copyright &y& Elsevier]

Abstract: Despite the well known advantages that a space-variant representation of the visual signal offers, the required adaptation of the algorithms developed in the Cartesian domain, before applying them in the log-polar space, has limited a wide use of such representation in visual processing applications. In this paper, we present a set of original rules for designing a discrete log-polar mapping that allows a direct application in the log-polar domain of the algorithms, based on spatial multi-scale and multi-orientation filtering, originally developed for the Cartesian domain. The advantage of the approach is to gain, without modifications, an effective space-variance and data reduction. Such design strategies are based on a quantitative analysis of the relationships between the spatial filtering and the space-variant representation. We assess the devised rules by using a distributed approach based on a bank of band-pass filters to compute reliable disparity maps, by providing quantitative measures of the computational load and of the accuracy of the computed visual features. [Copyright &y& Elsevier]

Abstract: Vanishing point detection algorithms based on 2D histogramming techniques have been employed in a variety of computer vision systems. Previous algorithms achieved some good results but still failed to maintain a balanced performance in both accuracy and time. Recent research () shows that, vanishing point detection could be converted to a 1D histogram search problem, which largely accelerates the procedure. In this paper, we further improve this idea and propose a complete scheme for vanishing point detection from images of the so called “Manhattan world”. We test our algorithm and some commonly used vanishing point detection methods on public database YorkUrbanDB and our own implemented database PKUCampusDB. Our algorithm shows significant performance improvements. [Copyright &y& Elsevier]

Abstract: New generations of telecommunications systems will include high-definition 3D video that provides a telepresence feeling. These systems require high-quality depth maps to be generated in a very short time (very low latency, typically about 40ms). Classical Belief Propagation algorithms (BP) generate high-quality depth maps but they require huge memory bandwidths that limit low-latency implementations of stereo-vision systems with high-definition images. This paper proposes a real-time (latency inferior to 40ms) high-definition (1280×720) stereo matching algorithm using Belief Propagation with good immersive feeling (80 disparity levels). There are two main contributions. The first is an improved BP algorithm with pixel classification that outperforms classical BP while reducing the number of memory accesses. The second is an adaptive message compression technique with a low performance penalty that greatly reduces the memory traffic. The combination of these techniques outperforms classical BP by about 6.0% while reducing the memory traffic by more than 90%. [Copyright &y& Elsevier]

Abstract: In this letter, we propose a robust, linear in time modification of Aktouf, Bertrand and Perroton’s algorithm for tunnel (3D hole) closing in 3D volumetric objects. Our algorithm is insensitive to small distortions and branches. The algorithm has been tested on various 3D images including very complicated 3D crack propagation images. The results of the tests, discussion of the algorithm properties and future research plans are also included in the paper. [Copyright &y& Elsevier]

Abstract: In this paper we propose a machine learning approach to classify melanocytic lesions as malignant or benign, using dermoscopic images. The lesion features used in the classification framework are inspired on border, texture, color and structures used in popular dermoscopy algorithms performed by clinicians by visual inspection. The main weakness of dermoscopy algorithms is the selection of a set of weights and thresholds, that appear not to be robust or independent of population. The use of machine learning techniques allows to overcome this issue. The proposed method is designed and tested on an image database composed of 655 images of melanocytic lesions: 544 benign lesions and 111 malignant melanoma. After an image pre-processing stage that includes hair removal filtering, each image is automatically segmented using well known image segmentation algorithms. Then, each lesion is characterized by a feature vector that contains shape, color and texture information, as well as local and global parameters. The detection of particular dermoscopic patterns associated with melanoma is also addressed, and its inclusion in the classification framework is discussed. The learning and classification stage is performed using AdaBoost with C4.5 decision trees. For the automatically segmented database, classification delivered a specificity of 77% for a sensitivity of 90%. The same classification procedure applied to images manually segmented by an experienced dermatologist yielded a specificity of 85% for a sensitivity of 90%. [Copyright &y& Elsevier]

Abstract: This paper presents a novel robust image alignment technique that performs joint geometric and photometric registration in the total least square (TLS) sense. Therefore, we employ the total least square metric instead of the ordinary least square (OLS) metric, which is commonly used in the literature. While the OLS model is sufficient to tackle geometric registration problems, it gives no mutually consistent estimates when dealing with photometric deformations. By introducing a new TLS model, we obtain mutually consistent parameters. Experimental results show that our method is indeed more consistent and accurate in presence of noise compared to existing joint registration algorithms. [Copyright &y& Elsevier]

Abstract: Discretization of continuous data is one of the important pre-processing tasks in data mining and knowledge discovery. Generally speaking, discretization can lead to improved predictive accuracy of induction algorithms, and the obtained rules are normally shorter and more understandable. In this paper, we present the Class-Attribute Coherence Maximization (CACM) algorithm and the Efficient-CACM algorithm. We have compared the performance of our algorithms with the most relevant discretization algorithm, Fast Class-Attribute Interdependence Maximization (Fast-CAIM) discertization algorithm (). Empirical evaluation of our algorithms and Fast-CAIM on 12 well-known datasets shows that ours generate the superior discretization scheme, which can significantly improve the classification performance of C4.5 and RBF-SVM classifier. As to the execution time of discretization, ours also prove faster than Fast-CAIM algorithm, with the Efficient-CACM algorithm having the shortest execution time. [Copyright &y& Elsevier]

Abstract: Citrus canker, a bacterial disease of citrus tree leaves, causes significant damage to citrus production worldwide. Effective and fast disease detection methods must be undertaken to minimize the losses of citrus canker infection. In this paper, we present a new approach based on global features and zone-based local features to detect citrus canker from leaf images collected in field which is more difficult than the leaf images captured in labs. Firstly, an improved AdaBoost algorithm is used to select the most significant features of citrus lesions for the segmentation of the lesions from their background. Then a canker lesion descriptor is proposed which combines both color and local texture distribution of canker lesion zones suggested by plant phytopathologists. A two-level hierarchical detection structure is developed to identify canker lesions. Thirdly, we evaluate the proposed method and its comparison with other approaches, and the experimental results show that the proposed approach achieves similar classification accuracy as human experts. [Copyright &y& Elsevier]

Abstract: The original local binary fitting (LBF) model is sensitive to contour initialization and thus easily obtains an inaccurate result due to improper initialization. This paper presents a new method that not only can arrive at sub-pixel accuracy, but also allows for more flexible initialization of the contour. Two important terms play main role in our new method. One is an image gradient alignment term (IGA) which uses the directional information of the image gradient, the other is a local intensity fitting term (LIF) which makes use of local region information. The integration of the above two terms prevents our method from being sensitive to contour initialization. In addition, a global intensity fitting term (GIF) multiplied by a stopping function is included, which can speed up our algorithm while do not influence the accuracy of the segmentation result. Using the simple central difference, the gradient descend flow equation for the level set function can be easily and efficiently implemented. The results on several synthetic and real images demonstrate the effectiveness and accuracy of our method. [Copyright &y& Elsevier]

Abstract: In this paper, we propose an improved two-stage scheme for the removal of salt-and-pepper noise based on an efficient impulse detector and the edge-preserving total variation inpainting model. We test the proposed algorithm on the four images corrupted by salt-and-pepper noise with a wide range of noise levels varied from 10% to 95%, and compare it with the six other well known filters. Further, we evaluate quantitatively the performance of these filters with the peak signal-to-noise ratio (PSNR), the measure of structural similarity (SSIM) and the image enhancement factor (IEF). The experimental results have demonstrated that the proposed filter performs impressively in noise suppression and edge preservation. [Copyright &y& Elsevier]

Abstract: This paper presents a fast simulated annealing framework for combining multiple clusterings based on agreement measures between partitions, which are originally used to evaluate a clustering algorithm. Although we can follow a greedy strategy to optimize these measures as the objective functions of clustering ensemble, it may suffer from local convergence and simultaneously incur too large computational cost. To avoid local optima, we consider a simulated annealing optimization scheme that operates through single label changes. Moreover, for the measures between partitions based on the relationship (joined or separated) of pairs of objects, we can update them incrementally for each label change, which ensures that our optimization scheme is computationally feasible. The experimental evaluations demonstrate that the proposed framework can achieve promising results. [Copyright &y& Elsevier]

Abstract: According to a certain misconception sometimes met in the literature: for the nearest-neighbors algorithms there is no fixed hypothesis class of limited Vapnik–Chervonenkis dimension. In the paper a simple reformulation (not a modification) of the nearest-neighbors algorithm is shown where instead of a natural number k, a percentage α ∈(0,1) of nearest neighbors is used. Owing to this reformulation one can construct sets of approximating functions, which we prove to have finite VC dimension. In a special (but practical) case this dimension is equal to ⌊2/α⌋. It is also then possible to form a sequence of sets of functions with increasing VC dimension, and to perform complexity selection via cross-validation or similarly to the structural risk minimization framework. Results of such experiments are also presented. [Copyright &y& Elsevier]

Abstract: This paper proposes new low-dimensional image features that enable images to be very efficiently matched. Image matching is one of the key technologies for many vision-based applications, including template matching, block motion estimation, video compression, stereo vision, image/video near-duplicate detection, similarity join for image/video database, and so on. Normalized cross correlation (NCC) is one of widely used method for image matching with preferable characteristics such as robustness to intensity offsets and contrast changes, but it is computationally expensive. The proposed features, derived by the method of Lagrange multipliers, can provide upper-bounds of NCC as a simple dot product between two low-dimensional feature vectors. By using the proposed features, NCC-based image matching can be effectively accelerated. The matching performance with the proposed features is demonstrated using an image database obtained from actual broadcast videos. The new features are shown to outperform other methods: multilevel successive elimination algorithm (MSEA), discrete cosine transform (DCT) coefficients, and histograms, achieving very high precision while only slightly sacrificing recall. [Copyright &y& Elsevier]

Abstract: Online learning algorithms have been preferred in many applications due to their ability to learn by the sequentially arriving data. One of the effective algorithms recently proposed for training single hidden-layer feedforward neural networks (SLFNs) is online sequential extreme learning machine (OS-ELM), which can learn data one-by-one or chunk-by-chunk at fixed or varying sizes. It is based on the ideas of extreme learning machine (ELM), in which the input weights and hidden layer biases are randomly chosen and then the output weights are determined by the pseudo-inverse operation. The learning speed of this algorithm is extremely high. However, it is not good to yield generalization models for noisy data and is difficult to initialize parameters in order to avoid singular and ill-posed problems. In this paper, we propose an improvement of OS-ELM based on the bi-objective optimization approach. It tries to minimize the empirical error and obtain small norm of network weight vector. Singular and ill-posed problems can be overcome by using the Tikhonov regularization. This approach is also able to learn data one-by-one or chunk-by-chunk. Experimental results show the better generalization performance of the proposed approach on benchmark datasets. [Copyright &y& Elsevier]

Abstract: Size Theory has proven to be a useful geometrical/topological approach to shape comparison. Originally introduced by considering 1-dimensional properties of shapes, described by means of real-valued functions, it has recently been generalized to taking into account multi-dimensional properties coded by functions valued in . This has led to the introduction of a shape descriptor called k-dimensional size function, and the k-dimensional matching distance to compare size functions. This paper presents new theoretical results about the 2-dimensional matching distance, leading to the formulation of an algorithm for its approximation up to an arbitrary error threshold. Experiments on 3D object comparison are shown to discuss the efficacy and effectiveness of the algorithm. [Copyright &y& Elsevier]

Abstract: In recent years, much attention has been given to the problem of outlier detection, whose aim is to detect outliers – objects who behave in an unexpected way or have abnormal properties. The identification of outliers is important for many applications such as intrusion detection, credit card fraud, criminal activities in electronic commerce, medical diagnosis and anti-terrorism, etc. In this paper, we propose a hybrid approach to outlier detection, which combines the opinions from boundary-based and distance-based methods for outlier detection (). We give a novel definition of outliers –BD (boundary and distance)-based outliers, by virtue of the notion of boundary region in rough set theory and the definitions of distance-based outliers. An algorithm to find such outliers is also given. And the effectiveness of our method for outlier detection is demonstrated on two publicly available databases. [Copyright &y& Elsevier]

Abstract: Given a segmentation result (an alpha matte or a binary mask) of the former frame, foreground prediction is a process of estimating the probability that each pixel in the current frame belongs to the foreground. It plays a very important role in bilayer segmentation of videos, especially videos with non-static backgrounds. In this paper, a new foreground prediction algorithm which is called opacity propagation is proposed. It can propagate the opacity values of the former frame to the current frame by minimizing a cost function that is constructed by assuming the spatiotemporally local color smoothness of the video. Optical flow and probability density estimation based on a local color model are employed to find the corresponding pixels of two adjacent frames. An OPSIC (opacity propagation with sudden illumination changes) algorithm is also proposed which is an improvement of our proposed opacity propagation algorithm because it adds a simple color transformation model. As far as we know, this is the first algorithm that can predict the foreground accurately when the illumination changes suddenly. The opacity map (OM) generated by the opacity propagation algorithm is usually more accurate than the previously used probability map (PM). The experiments demonstrate the effectiveness of our algorithm. [Copyright &y& Elsevier]

Abstract: Clustering is one of the widely used knowledge discovery techniques to reveal structures in a dataset that can be extremely useful to the analyst. In iterative clustering algorithms the procedure adopted for choosing initial cluster centers is extremely important as it has a direct impact on the formation of final clusters. Since clusters are separated groups in a feature space, it is desirable to select initial centers which are well separated. In this paper, we have proposed an algorithm to compute initial cluster centers for k-means algorithm. The algorithm is applied to several different datasets in different dimension for illustrative purposes. It is observed that the newly proposed algorithm has good performance to obtain the initial cluster centers for the k-means algorithm. [Copyright &y& Elsevier]

Abstract: This paper introduces a new method for the simultaneous computation of sets of lines meeting at multiple vanishing points through the use of the Expectation–Maximisation (EM) algorithm. The proposed method is based on the formulation of novel error functions in the projective plane between lines and points which involve the use of non-linear optimisation procedures that allow to treat equally finite and infinite vanishing points. These functions are included into the EM framework, which handles the multi-modality of the problem naturally with mixture models. Results show that the proposed method of joint computation of vanishing points and lines converging at such points enhances the robustness and the accuracy of locating these points. [Copyright &y& Elsevier]

Abstract: The classical notion of clustering is to induce an equivalence class partition on a set of points, each class, being a homogeneous group, is called a cluster. Since it is an equivalence class partition, a point must belong to one and exactly one cluster. However in many applications, data distributions are such that only a subset of the points tends to flock under some distinct clusters while others go random. The present paper introduces an algorithm to find an optimal subset of points (ideally filtering out the random ones) with sufficient grouping tendency. It builds the neighborhood population around every point and picks up top k dense regions with possible reshuffling of points in post-processing. Performance of the algorithm is evaluated with applications onto real and simulated data. Comparative analysis on different quality indices with some other state-of-the-art algorithms establishes effectiveness of the approach. [Copyright &y& Elsevier]

Abstract: For multimodality images, a novel fusion algorithm based on the shiftable complex directional pyramid transform (SCDPT) is proposed in this paper. As well, with the aid of the structural similarity (SSIM) index, a ‘similarity-based’ idea is employed to distinguish regions with ‘redundant’ or ‘complementary’ information between source imagers before the SCDPT coefficients are merged. A ‘weighted averaging’ scheme for regions with ‘redundant’ information and a ‘selecting’ scheme for regions with ‘complementary’ information are then employed, respectively. When merging the low-pass subband coefficients, the SSIM index in spatial domain (SP-SSIM) is employed as similarity measure, and three types of regions are thus determined. Especially, for regions with similar intensity values but different intensity changing directions between source images, a ‘selecting’ scheme based on gradient and energy is proposed. When merging the directional band-pass subband coefficients, the SSIM index in complex wavelet domain (CW-SSIM) is employed as similarity measure. With the CW-SSIM index, not only the magnitude information but also the phase information of SCDPT coefficients can be employed. Compared to the traditional energy matching (EM) index based fusion methods, the proposed method can better deal with ‘redundant’ and ‘complementary’ information of source images. In addition, because of the shift-invariance of the SCDPT and the CW-SSIM index, the proposed fusion algorithm performs well even if the input images are not well registered. Several sets of experimental results demonstrate the validity and feasibility of the proposed method in terms of both visual quality and objective evaluation. [Copyright &y& Elsevier]

Abstract: The present paper considers the problem of partitioning a dataset into a known number of clusters using the sum of squared errors criterion (SSE). A new clustering method, called DE-KM, which combines differential evolution algorithm (DE) with the well known K-means procedure is described. In the method, the K-means algorithm is used to fine-tune each candidate solution obtained by mutation and crossover operators of DE. Additionally, a reordering procedure which allows the evolutionary algorithm to tackle the redundant representation problem is proposed. The performance of the DE-KM clustering method is compared to the performance of differential evolution, global K-means method, genetic K-means algorithm and two variants of the K-means algorithm. The experimental results show that if the number of clusters K is sufficiently large, DE-KM obtains solutions with lower SSE values than the other five algorithms. [Copyright &y& Elsevier]

Abstract: Clustering is one of the most important unsupervised learning problems and it consists of finding a common structure in a collection of unlabeled data. However, due to the ill-posed nature of the problem, different runs of the same clustering algorithm applied to the same data-set usually produce different solutions. In this scenario choosing a single solution is quite arbitrary. On the other hand, in many applications the problem of multiple solutions becomes intractable, hence it is often more desirable to provide a limited group of “good” clusterings rather than a single solution. In the present paper we propose the least squares consensus clustering. This technique allows to extrapolate a small number of different clustering solutions from an initial (large) ensemble obtained by applying any clustering algorithm to a given data-set. We also define a measure of quality and present a graphical visualization of each consensus clustering to make immediately interpretable the strength of the consensus. We have carried out several numerical experiments both on synthetic and real data-sets to illustrate the proposed methodology. [Copyright &y& Elsevier]

Abstract: Non-Gaussian triplet Markov random fields (TMF) model is suitable for dealing with multi-class segmentation of nonstationary and non-Gaussian synthetic aperture radar (SAR) images. However, the segmentation of SAR images utilizing this model still fails to resolve the misclassifications due to the inaccuracy of edge location. In this paper, we propose a new unsupervised multi-class segmentation algorithm by fusing the traditional energy function of TMF model with the principle of edge penalty. Through the introduction of the penalty function based on local edge strength information, the new energy function could prevent segment from smoothing across boundaries. Then we optimize the objective function that stems from the new energy function to obtain an iterative multi-region merging Bayesian maximum posterior mode (MPM) segmentation equation for the new segmentation algorithm. The effectiveness of the proposed algorithm is demonstrated by application to simulated data and real SAR images. [Copyright &y& Elsevier]

Abstract: Recent years have witnessed great success of manifold learning methods in understanding the structure of multidimensional patterns. However, most of these methods operate in a batch mode and cannot be effectively applied when data are collected sequentially. In this paper, we propose a general incremental learning framework, capable of dealing with one or more new samples each time, for the so-called spectral embedding methods. In the proposed framework, the incremental dimensionality reduction problem reduces to an incremental eigen-problem of matrices. Furthermore, we present, using this framework as a tool, an incremental version of Hessian eigenmaps, the IHLLE method. Finally, we show several experimental results on both synthetic and real world datasets, demonstrating the efficiency and accuracy of the proposed algorithm. [Copyright &y& Elsevier]