Search

Abstract: Many parametric image alignment approaches assume equality of the images to register up to motion compensation. In presence of noise this assumption does not hold. In particular, for gradient-based approaches, which rely on the optimization of an error functional with gradient descent methods, the performances depend on the amount of noise in each image. We propose in this paper to use the Asymmetric Composition on Lie groups (ACL) formulation of the alignment problem to improve the robustness in presence of asymmetric levels of noise. The ACL formulation, generalizing state-of-the-art gradient-based image alignment, introduces a parameter to weight the influence of the images during the optimization. Three new methods are presented to estimate this asymmetry parameter: one supervised (MVACL) and two fully automatic (AACL and GACL). Theoretical results and experimental validation show how the new algorithms improve robustness in presence of noise. Finally, we illustrate the interest of the new approaches for object tracking under low-light conditions. [Copyright &y& Elsevier]

Abstract: Traditional clustering ensemble methods combine all obtained clustering results at hand. However, we can often achieve a better clustering solution if only parts of the clustering results available are combined. In this paper, we generalize the selective clustering ensemble algorithm proposed by Azimi and Fern and a novel clustering ensemble method, SELective Spectral Clustering Ensemble (SELSCE), is proposed. The component clusterings of the ensemble system are generated by spectral clustering (SC) capable of engendering diverse committees. The random scaling parameter, Nyström approximation are used to perturb SC for producing the components of the ensemble system. After the generation of component clusterings, the bagging technique, usually applied in supervised learning, is used to assess the component clustering. We randomly pick part of the available clusterings to get a consensus result and then compute normalized mutual information (NMI) or adjusted rand index (ARI) between the consensus result and the component clusterings. Finally, the components are ranked by aggregating multiple NMI or ARI values. The experimental results on UCI dataset and images demonstrate that the proposed algorithm can achieve a better result than the traditional clustering ensemble methods. [Copyright &y& Elsevier]

Abstract: The appearance of a face image is severely affected by illumination conditions that will hinder the automatic face recognition process. To recognize faces under varying lighting conditions, a homomorphic filtering-based illumination normalization method is proposed in this paper. In this work, the effect of illumination is effectively reduced by a modified implementation of homomorphic filtering whose key component is a Difference of Gaussian (DoG) filter, and the contrast is enhanced by histogram equalization. The resulted face image is not only reduced illumination effect but also preserved edges and details that will facilitate the further face recognition task. Among others, our method has the following advantages: (1) neither does it need any prior information of 3D shape or light sources, nor many training samples thus can be directly applied to single training image per person condition; and (2) it is simple and computationally fast because there are mature and fast algorithms for the Fourier transform used in homomorphic filter. The Eigenfaces method is chosen to recognize the normalized face images. Experimental results on the Yale face database B and the CMU PIE face database demonstrate the significant performance improvement of the proposed method in the face recognition system for the face images with large illumination variations. [Copyright &y& Elsevier]

Abstract: Text categorization presents unique challenges to traditional classification methods due to the large number of features inherent in the datasets from real-world applications of text categorization, and a great deal of training samples. In high-dimensional document data, the classes are typically categorized only by subsets of features, which are typically different for the classes of different topics. This paper presents a simple but effective classifier for text categorization using class-dependent projection based method. By projecting onto a set of individual subspaces, the samples belonging to different document classes are separated such that they are easily to be classified. This is achieved by developing a new supervised feature weighting algorithm to learn the optimized subspaces for all the document classes. The experiments carried out on common benchmarking corpuses showed that the proposed method achieved both higher classification accuracy and lower computational costs than some distinguishing classifiers in text categorization, especially for datasets including document categories with overlapping topics. [Copyright &y& Elsevier]

Abstract: K-hyperline clustering is an iterative algorithm based on singular value decomposition and it has been successfully used in sparse component analysis. In this paper, we prove that the algorithm converges to a locally optimal solution for a given set of training data, based on Lloyd’s optimality conditions. Furthermore, the local optimality is shown by developing an Expectation-Maximization procedure for learning dictionaries to be used in sparse representations and by deriving the clustering algorithm as its special case. The cluster centroids obtained from the algorithm are proved to tessellate the space into convex Voronoi regions. The stability of clustering is shown by posing the problem as an empirical risk minimization procedure over a function class. It is proved that, under certain conditions, the cluster centroids learned from two sets of i.i.d. training samples drawn from the same probability space become arbitrarily close to each other, as the number of training samples increase asymptotically. [Copyright &y& Elsevier]

Abstract: In this paper, we propose a novel algorithm for object template tracking and its drift correction. It can prevent the tracking drift effectively, and save the time of an additional correction tracking. In our algorithm, the total energy function consists of two terms: the tracking term and the drift correction term. We minimize the total energy function synchronously for template tracking and weighted active drift correction. The minimization of the active drift correction term is achieved by the inverse compositional algorithm with a weighted L2 norm, which is incorporated into traditional affine image alignment (AIA) algorithm. Its weights can be adaptively updated for each template. For diminishing the accumulative error in tracking, we design a new template update strategy that chooses a new template with the lowest matching error. Finally, we will present various experimental results that validate our algorithm. These results also show that our algorithm achieves better performance than the inverse compositional algorithm for drift correction. [Copyright &y& Elsevier]

Abstract: In high-dimensional data, clusters of objects usually exist in subspaces; besides, different clusters probably have different shape volumes. Most existing methods for high-dimensional data clustering, however, only consider the former factor. They ignore the latter factor by assuming the same shape volume value for different clusters. In this paper we propose a new Gaussian mixture model (GMM) type algorithm for discovering clusters with various shape volumes in subspaces. We extend the GMM clustering method to calculate a local weight vector as well as a local variance within each cluster, and use the weight and variance values to capture main properties that discriminate different clusters, including subsets of relevant dimensions and shape volumes. This is achieved by introducing negative entropy of weight vectors, along with adaptively-chosen coefficients, into the objective function of the extended GMM. Experimental results on both synthetic and real datasets show that the proposed algorithm outperforms its competitors, especially when applying to high-dimensional datasets. [Copyright &y& Elsevier]

Abstract: In this paper a new approach is presented for tackling the problem of identifying the author of a handwritten text. This problem is solved with a simple, yet powerful, modification of the so called ALVOT family of supervised classification algorithms with a novel differentiated-weighting scheme. Compared to other previously published approaches, the proposed method significantly reduces the number and complexity of the text-features to be extracted from the text. Also, the specific combination of line-level and word-level features used introduces an eclectic paradigm between texture-related and structure-related approaches. [Copyright &y& Elsevier]

Abstract: A new approach for edge detection using a combination of bacterial foraging algorithm (BFA) and probabilistic derivative technique derived from Ant Colony Systems, is presented in this paper. The foraging behavior of some species of bacteria like Escherichia coli can be hypothetically modeled as an optimization process. A group of bacteria search for nutrients in a way that maximizes the energy obtained per unit time spent during the foraging. The proposed approach aims at driving the bacteria through the edge pixels. The direction of movement of the bacteria is found using a direction probability matrix, computed using derivatives along the possible directions. Rules defining the derivatives are devised to ensure that the variation of intensity due to noise is discarded. Quantitative analysis of the feasibility of the proposed approach and its comparison with other standard edge detection operators in terms of kappa and entropy are given. The effect of initial values of parameters of BFA on the edge detection is discussed. [Copyright &y& Elsevier]

Abstract: This paper proposes robust refinement methods to improve the popular patch multi-view 3D reconstruction algorithm by . Specifically, a new method is proposed to improve the robustness by removing outliers based on a filtering approach. In addition, this work also proposes a method to divide the 3D points in to several buckets for applying the sparse bundle adjustment algorithm (SBA) individually, removing the outliers and finally merging them. The residuals are used to filter potential outliers to reduce the re-projection error used as the performance evaluation of refinement. In our experiments, the original mean re-projection error is about 47.6. After applying the proposed methods, the mean error is reduced to 2.13. [Copyright &y& Elsevier]

Abstract: Combinatorial maps explicitly encode orientations of edges around vertices, and have been used in many fields. In this paper, we address the problem of searching for patterns in model maps by putting forward the concept of symbol graph. A symbol graph will be constructed and stored for each model map in the preprocessing. Furthermore, an algorithm for submap isomorphism is presented based on symbol sequence searching in the symbol graphs. The computational complexity of this algorithm is quadratic in the worst case if we neglect the preprocessing step. [Copyright &y& Elsevier]

Abstract: Recognition of handwritten Arabic cursive texts is a complex task due to the similarities between letters under different writing styles. In this paper, a word-based off-line recognition system is proposed, using Hidden Markov Models (HMMs). The method employed involves three stages, namely preprocessing, feature extraction and classification. First, words from input scripts are segmented and normalized. Then, a set of intensity features are extracted from each of the segmented words, which is based on a sliding window moving across each mirrored word image. Meanwhile, structure-like features are also extracted including number of subwords and diacritical marks. Finally, these features are applied in a combined scheme for classification. Intensity features are used to train a HMM classifier, whose results are re-ranked using structure-like features for improved recognition rate. In order to validate the proposed techniques, extensive experiments were carried out using the IFN/ENIT database which contains 32,492 handwritten Arabic words. The proposed algorithm yields superior results of improved accuracy in comparison with several typical methods. [Copyright &y& Elsevier]

Abstract: In this paper, we proposed an adaptive pixon represented segmentation (APRS) algorithm for 3D magnetic resonance (MR) brain images. Different from traditional method, an adaptive mean shift algorithm was adopted to adaptively smooth the query image and create a pixon-based image representation. Then K-means algorithm was employed to provide an initial segmentation by classifying the pixons in image into a predefined number of tissue classes. By using this segmentation as initialization, expectation-maximization (EM) iterations composed of bias correction, a priori digital brain atlas information, and Markov random field (MRF) segmentation were processed. Pixons were assigned with final labels when the algorithm converges. The adoption of bias correction and brain atlas made the current method more suitable for brain image segmentation than the previous pixon based segmentation algorithm. The proposed method was validated on both simulated normal brain images from BrainWeb and real brain images from the IBSR public dataset. Compared with some other popular MRI segmentation methods, the proposed method exhibited a higher degree of accuracy in segmenting both simulated and real 3D MRI brain data. The experimental results were numerically assessed using Dice and Tanimoto coefficients. [Copyright &y& Elsevier]

Abstract: Automatic image segmentation is always a fundamental but challenging problem in computer vision. The simplest approach to image segmentation may be clustering feature vectors of pixels at first, then labeling each pixel with its corresponding cluster. This requires that the clustering on feature space must be robust. However, most of popular clustering algorithms could not obtain a robust clustering result yet, if the clusters in feature space have a complex distribution. Generally, for most of clustering-based segmentation methods, it still needs more constraints of positional relations between pixels in image lattice to be utilized during the procedure of clustering. Our works in this paper address the problem of image segmentation under the paradigm of pure clustering-then-labeling. A robust clustering algorithm which could maintain good coherence of data in feature space is proposed and utilized to do clustering on the L∗a∗b∗ color feature space of pixels. Image segmentation is straightforwardly obtained by setting each pixel with its corresponding cluster. Further, based on the theory of Minimum Description Length, an effective approach to automatic parameter selection for our segmentation method is also proposed. We test our segmentation method on Berkeley segmentation database, and the experimental results show that our method compares favorably against some state-of-the-art segmentation methods. [Copyright &y& Elsevier]

Abstract: This paper proposes a generalized method for playfield segmentation of various sport videos. It first estimates the probability density function (pdf) of color components of an image frame. Two hill-climbing schemes, which employ two-dimensional pdf and one-dimensional pdf, respectively, are proposed for clustering. Next, a novel algorithm that utilizes the domain knowledge of sport playfields is developed to merge those clusters into four color classes: red, green, blue, and gray. Finally, a simple scheme fuses small regions into their adjacent large regions to obtain the segmentation result. The experimental results indicate that the proposed method effectively segments the playfield regions of various sport videos. [Copyright &y& Elsevier]

Abstract: Accurate and efficient automatic or semi-automatic brain image segmentation methods are of great interest to both scientific and clinical researchers of the human central neural system. Cerebral white matter segmentation in brain Magnetic Resonance Imaging (MRI) data becomes a challenging problem due to a combination of several factors like low contrast, presence of noise and imaging artifacts, partial volume effects, intrinsic tissue variation due to neurodevelopment and neuropathologies, and the highly convoluted geometry of the cortex. In this paper, we propose a new set of edge weights for the traditional graph cut algorithm () to correctly segment the cerebral white matter from T1-weighted MRI sequence. In this algorithm, the edge weights of are modified by comparing the probabilities of an individual voxel and its neighboring voxels to belong to different segmentation classes. A shape prior in form of a series of ellipses is used next to model the contours of the human skull in various 2D slices in the sequence. This shape constraint is imposed to prune the original graph constructed from the input to form a subgraph consisting of voxels within the skull contours. Our graph cut algorithm with new set of edge weights is applied to the above subgraph, thereby increasing the segmentation accuracy as well as decreasing the computation time. Average segmentation errors for the proposed algorithm, the graph cut algorithm (), and the Expectation Maximization Segmentation (EMS) algorithm in terms of Dice coefficients are found to be (3.72±1.12)%, (14.88±1.69)%, and (11.95±5.2)%, respectively. [Copyright &y& Elsevier]

Abstract: This paper reviews the TPS-RPM algorithm () for robustly registering two sets of points and demonstrates from a theoretical point of view its inherent limited performance when outliers are present in both point sets simultaneously. A double-sided outlier handling approach is proposed to overcome this limitation with a rigorous mathematical proof as the underlying theoretical support. This double-sided outlier handling approach is proved to be equivalent to the original formulation of the point matching problem. For a practical application, we also extend the TPS-RPM algorithms to non-rigid image registration by registering two sets of sparse features extracted from images. The intensity information of the extracted features are incorporated into feature matching in order to reduce the impact from outliers. Our experiments demonstrate the double-sided outlier handling approach and the efficiency of intensity information in assisting outlier detection. [Copyright &y& Elsevier]

Abstract: Artificial neural networks techniques have been successfully applied in vector quantization (VQ) encoding. The objective of VQ is to statistically preserve the topological relationships existing in a data set and to project the data to a lattice of lower dimensions, for visualization, compression, storage, or transmission purposes. However, one of the major drawbacks in the application of artificial neural networks is the difficulty to properly specify the structure of the lattice that best preserves the topology of the data. To overcome this problem, in this paper we introduce merging algorithms for machine-fusion, boosting-fusion-based and hybrid-fusion ensembles of SOM, NG and GSOM networks. In these ensembles not the output signals of the base learners are combined, but their architectures are properly merged. We empirically show the quality and robustness of the topological representation of our proposed algorithm using both synthetic and real benchmarks datasets. [Copyright &y& Elsevier]

Abstract: Image representation by using polynomial moments is an interesting theme. In this paper, image representation by using orthogonal Gegenbauer function is presented. A novel method for accurate and fast computation of orthogonal Gegenbauer moments is proposed. The accurate values of Gegenbauer moments are obtained by mathematically integrating Gegenbauer polynomials multiplied by their weight functions over the digital image pixels. A novel recurrence formula is derived for the kernel generation. The proposed method removes the numerical approximation errors involved in conventional method. A fast algorithm is proposed to accelerate the moment’s computations. A comparison with the conventional method is performed. The obtained results explain the efficiency and the superiority of the proposed method. [Copyright &y& Elsevier]

Abstract: Corner matching in image sequences is an important and difficult problem that serves as a building block of several important applications of stereo vision etc. Normally, in area-based corner matching techniques, the linear measures like standard cross correlation coefficient, zero-mean (normalized) cross correlation coefficient, sum of absolute difference and sum of squared difference are used. Fuzzy logic is a powerful tool to solve many image processing problems because of its ability to deal with ambiguous data. In this paper, we use a similarity measure based on fuzzy correlations in order to establish the corner correspondence between sequence images in the presence of intensity variations and motion blur. The matching approach proposed here needs only to extract one set of corner points as candidates from the left image (first frame), and the positions of which in the right image (second frame) are determined by matching, not by extracting. Experiments conducted with the help of various sequences of images prove the superiority of our algorithm over standard and zero-mean cross correlation as well as one contemporary work using mutual information as a window similarity measure combined with graph matching techniques under non-ideal conditions. [Copyright &y& Elsevier]

Abstract: In this paper, an efficient feature extraction algorithm called orthogonal local spline discriminant projection (O-LSDP) is proposed for face recognition. Derived from local spline embedding (LSE), O-LSDP not only inherits the advantages of LSE which uses local tangent space as a representation of the local geometry so as to preserve the local structure, but also makes full use of class information and orthogonal subspace to improve discriminant power. Extensive experiments on several standard face databases demonstrate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Abstract: In this paper, the RGB ratio is defined according to a reference color so that an image can be transformed from a conventional color space to the RGB ratio space. Different to traditional distance measurement, a road color model is determined by an ellipse area in the RGB ratio space enclosed by the estimated boundaries. The proposed dynamic fuzzy logic, where fuzzy membership functions are defined according to estimated boundaries, is introduced to implement clustering rules, such that each pixel will have its own fuzzy membership function corresponding to its intensity. A basic neural network is trained and used to achieve parameter optimization. Experimental results for road detection demonstrate the robustness of the proposed approach to variations in intensity. To provide obstacle information, especially for visually impaired people, Musical Instrument Digital Interface (MIDI) is introduced as the sound generator, and image-to-MIDI mapping algorithm is proposed. Experimental results show that the proposed method can adapt to various road types, and the resulting audio information successfully indicates the position and size of obstacles. [ABSTRACT FROM AUTHOR]

Abstract: Text representation is a necessary procedure for text categorization tasks. Currently, bag of words (BOW) is the most widely used text representation method but it suffers from two drawbacks. First, the quantity of words is huge; second, it is not feasible to calculate the relationship between words. Semantic analysis (SA) techniques help BOW overcome these two drawbacks by interpreting words and documents in a space of concepts. However, existing SA techniques are not designed for text categorization and often incur huge computing cost. This paper proposes a concise semantic analysis (CSA) technique for text categorization tasks. CSA extracts a few concepts from category labels and then implements concise interpretation on words and documents. These concepts are small in quantity and great in generality and tightly related to the category labels. Therefore, CSA preserves necessary information for classifiers with very low computing cost. To evaluate CSA, experiments on three data sets (Reuters-21578, 20-NewsGroup and Tancorp) were conducted and the results show that CSA reaches a comparable micro- and macro-F 1 performance with BOW, if not better one. Experiments also show that CSA helps dimension sensitive learning algorithms such as k-nearest neighbor (kNN) to eliminate the “Curse of Dimensionality” and as a result reaches a comparable performance with support vector machine (SVM) in text categorization applications. In addition, CSA is language independent and performs equally well both in Chinese and English. [ABSTRACT FROM AUTHOR]

Abstract: In this paper, a novel local matching method called structure-preserved projections (SPP) is proposed for face recognition. Unlike most existing local matching methods which neglect the interactions of different sub-pattern sets during feature extraction, i.e., they assume different sub-pattern sets are independent; SPP takes the holistic context of the face into account and can preserve the configural structure of each face image in subspace. Moreover, the intrinsic manifold structure of the sub-pattern sets can also be preserved in our method. With SPP, all sub-patterns partitioned from the original face images are trained to obtain a unified subspace, in which recognition can be performed. The efficiency of the proposed algorithm is demonstrated by extensive experiments on three standard face databases (Yale, Extended YaleB and PIE). Experimental results show that SPP outperforms other holistic and local matching methods. [ABSTRACT FROM AUTHOR]

Abstract: Computed tomographic (CT) colonography is a promising alternative to traditional invasive colonoscopic methods used in the detection and removal of cancerous growths, or polyps in the colon. Existing computer-aided diagnosis (CAD) algorithms used in CT colonography typically employ the use of a classifier to discriminate between true and false positives generated by a polyp candidate detection system based on a set of features extracted from the candidates. However, these classifiers often suffer from a phenomenon termed the curse of dimensionality, whereby there is a marked degradation in the performance of a classifier as the number of features used in the classifier is increased. In addition, an increase in the number of features used also contributes to an increase in computational complexity and demands on storage space. This paper investigates the benefits of feature selection on a polyp candidate database, with the aim of increasing specificity while preserving sensitivity. Two new mutual information methods for feature selection are proposed in order to select a subset of features for optimum performance. Initial results show that the performance of the widely used support vector machine (SVM) classifier is indeed better with the use of a small set of features, with receiver operating characteristic curve (AUC) measures reaching 0.78–0.88. [ABSTRACT FROM AUTHOR]

Abstract: This paper examines the Rocchio algorithm and its application in text categorization. Existing approaches using global parameters optimization of Rocchio algorithm result in choosing one fixed prototype representing each category for multi-category text categorization problems. Therefore, they have limited discriminating power on different category’s distribution and their parameter optimization methods are based on weak representation ability of the negative samples consisting of several categories. We present a pairwise optimized Rocchio algorithm, which dynamically adjusts the prototype position between pairs of categories. Experiments were conducted on three benchmark corpora, the 20-Newsgroup, Reuters-21578 and TDT2. The results confirm that our proposed pairwise method achieves encouraging performance improvement over the conventional Rocchio method. A comparative study with the top notch text classifier Support Vector Machine (SVM) also shows the pairwise Rocchio method achieves competitive results. [ABSTRACT FROM AUTHOR]

Abstract: In this paper, a new and automatic moving-edge detection algorithm is proposed, based on using the heat flow analogy. This algorithm starts with anisotropic heat diffusion in the spatial domain, to remove noise and sharpen region boundaries for the purpose of obtaining high quality edge data. Then, isotropic and linear heat diffusion is applied in the temporal domain to calculate the total amount of heat flow. The moving-edges are represented as the total amount of heat flow out from the reference frame. The overall process is completed by non-maxima suppression and hysteresis thresholding to obtain binary moving-edges. Evaluation, on a variety of data, indicates that this approach can handle noise in the temporal domain because of the averaging inherent of isotropic heat flow. Results also show that this technique can detect moving-edges in image sequences, without background image subtraction. [ABSTRACT FROM AUTHOR]

Abstract: We address the balanced clustering problem where cluster sizes are regularized with submodular functions. The objective function for balanced clustering is a submodular fractional function, i.e., the ratio of two submodular functions, and thus includes the well-known ratio cuts as special cases. In this paper, we present a novel algorithm for minimizing this objective function (submodular fractional programming) using recent submodular optimization techniques. The main idea is to utilize an algorithm to minimize the difference of two submodular functions, combined with the discrete Newton method. Thus, it can be applied to the objective function involving any submodular functions in both the numerator and the denominator, which enables us to design flexible clustering setups. We also give theoretical analysis on the algorithm, and evaluate the performance through comparative experiments with conventional algorithms by artificial and real datasets. [ABSTRACT FROM AUTHOR]

Abstract: Similarity measurement is crucial to the performance of spectral clustering. The Gaussian kernel function is usually adopted as the similarity measure. However, with a fixed kernel parameter, the similarity between two data points is only determined by their Euclidean distance, and is not adaptive to their surroundings. In this paper, a local density adaptive similarity measure is proposed, which uses the local density between two data points to scale the Gaussian kernel function. The proposed similarity measure satisfies the clustering assumption and has an effect of amplifying intra-cluster similarity, thus making the affinity matrix clearly block diagonal. Experimental results on both synthetic and real world data sets show that the spectral clustering algorithm with our local density adaptive similarity measure outperforms the traditional spectral clustering algorithm, the path-based spectral clustering algorithm and the self-tuning spectral clustering algorithm. [ABSTRACT FROM AUTHOR]

Abstract: This paper presents a novel algorithm of fingerprint encryption which transforms fingerprint minutiae and performs matching in the transformed form. If an encrypted template is compromised, it can be cancelled by choosing just another transformed template. In our algorithm, a circular region is constructed around each minutia and non-invertible transformation is applied to all regions but only transformed regions are stored in the database. The proposed algorithm improves the accuracy of verification compared to related fuzzy vault systems. Experimental results show the comparative performance of matching using both transformed and original data. We find that transformed matching also has an impressive accuracy and speed. [ABSTRACT FROM AUTHOR]

Abstract: The Information Bottleneck principle provides a systematic method to extract relevant features from complex data sets, and it models features extraction as data compression and quantifies the relevance of extracted feature by how much information it preserved about a specified feature. How to construct an optimal solution to IB remains a problem. The current Information Bottleneck (IB) algorithms only utilize the information between element pairs, and ignore the information among the neighborhood of elements. This is one of the major reasons for most IB algorithms’ failure to preserve as much relative information as possible, which further limits IB applicability in many areas. In this paper, we present the concept of density connectivity component, by which the information loss among the neighbors of an element, rather than the information loss between paired elements, can be considered. Then, we introduce this concept into the current agglomerative IB algorithm (aIB) and sequential IB algorithm (sIB), and propose two density-based IB algorithms, DaIB and DsIB. The experiment results on the benchmark data sets indicate that the DaIB and DsIB algorithm can preserve more relevant information and achieve higher precision than the aIB and sIB algorithm, respectively. [ABSTRACT FROM AUTHOR]

Abstract: Principal component analysis (PCA) is widely used in recently proposed manifold learning algorithms to provide approximate local tangent spaces. However, such approximations provided by PCA may be inaccurate when local neighborhoods of the data manifold do not lie in or close to a linear subspace. Furthermore, the approximated tangent spaces can not fit the change in data distribution density. In this paper, a new method is proposed for providing faithful approximations to the local tangent spaces of a data manifold, which is proved to be more accurate than PCA. With this new method, an improved local tangent space alignment (ILTSA) algorithm is developed, which can efficiently recover the geometric structure of data manifolds even in the case when data are sparse or non-uniformly distributed. Experimental results are presented to illustrate the better performance of ILTSA on both synthetic data and image data. [ABSTRACT FROM AUTHOR]

Abstract: Skeletons have been playing an important role in shape analysis since the introduction of the medial axis in the sixties. The original medial axis definition is in the continuous Euclidean space. It is a skeleton with the following characteristics: centered, thin, homotopic (it has the same topology as the object), and reversible (sufficient for the reconstruction of the object). The discrete version of the Euclidean medial axis (MA) is also reversible and centered, but no longer homotopic nor thin. The combination of the MA with homotopic thinning algorithms can result in a topology preserving, reversible skeleton. However, such combination may generate thicker skeletons, and the choice of the thinning algorithm is not obvious. A robust and well defined framework for fast homotopic thinning available in the literature is defined in the domain of abstract complexes. Since the abstract complexes are represented in a doubled resolution grid, some authors have extended the MA to a doubled resolution grid and defined the discrete Euclidean medial axis in higher resolution (HMA). The HMA can be combined with the thinning framework defined on abstract complexes. Other authors have given an alternative definition of medial axis, which is a subset of the MA, called reduced discrete medial axis (RDMA). The RDMA is reversible, thinner than the MA, and it can be computed in optimal time. In this paper, we extend the RDMA to the doubled resolution grid and we define the high-resolution RDMA (HRDMA). We provide an optimal algorithm to compute the HRDMA. The HRDMA can be combined with the thinning framework defined on abstract complexes. The skeleton obtained by such combination is provided with strong characteristics, not found in the literature. [Copyright &y& Elsevier]

Abstract: The fast evolution of digital video has brought many new multimedia applications and, as a consequence, has increased the amount of research into new technologies that aim at improving the effectiveness and efficiency of video acquisition, archiving, cataloging and indexing, as well as increasing the usability of stored videos. Among possible research areas, video summarization is an important topic that potentially enables faster browsing of large video collections and also more efficient content indexing and access. Essentially, this research area consists of automatically generating a short summary of a video, which can either be a static summary or a dynamic summary. In this paper, we present VSUMM, a methodology for the production of static video summaries. The method is based on color feature extraction from video frames and k-means clustering algorithm. As an additional contribution, we also develop a novel approach for the evaluation of video static summaries. In this evaluation methodology, video summaries are manually created by users. Then, several user-created summaries are compared both to our approach and also to a number of different techniques in the literature. Experimental results show – with a confidence level of 98% – that the proposed solution provided static video summaries with superior quality relative to the approaches to which it was compared. [Copyright &y& Elsevier]

Abstract: Marginal information is of great importance for classification. This paper presents a new nonparametric linear discriminant analysis method named Push–Pull marginal discriminant analysis (PPMDA), which takes full advantage of marginal information. For two-class cases, the idea of this method is to determine projected directions such that the marginal samples of one class are pushed away from the between-class marginal samples as far as possible and simultaneously pulled to the within-class samples as close as possible. This idea can be extended for multi-class cases and give rise to the PPMDA algorithm for feature extraction of multi-class problems. The proposed method is evaluated using the CENPARMI handwritten numeral database, the Extended Yale face database B and the ORL database. Experimental results show the effectiveness of the proposed method and its advantage after performance over the state-of-the-art feature extraction methods. [Copyright &y& Elsevier]

Abstract: In recent year, intelligent visual surveillance has become more and more important for enhanced security. In this paper, we will introduce some recent work in image and video understanding. First, we will give an introduction of the related video surveillance system in recent years, in particular, we will describe algorithms and systems developed in our group for the automatic interpretation of human and vehicle motion in surveillance videos, where automatic interpretation involves object motion detection, object classification and recognition, object tracking and the analysis of object behaviors in order to detect abnormal behaviors. We also give some examples of the real applications of Vs-star system. [Copyright &y& Elsevier]

Abstract: We consider the problem of fitting linearly parameterized models, that arises in many computer vision problems such as road scene analysis. Data extracted from images usually contain non-Gaussian noise and outliers, which makes non-robust estimation methods ineffective. In this paper, we propose an overview of a Lagrangian formulation of the Half-Quadratic approach by, first, revisiting the derivation of the well-known Iterative Re-weighted Least Squares (IRLS) robust estimation algorithm. Then, it is shown that this formulation helps derive the so-called Modified Residuals Least Squares (MRLS) algorithm. In this framework, moreover, standard theoretical results from constrained optimization can be invoked to derive convergence proofs easier. The interest of using the Lagrangian framework is also illustrated by the extension to the problem of the robust estimation of sets of linearly parameterized curves, and to the problem of robust fitting of linearly parameterized regions. To demonstrate the relevance of the proposed algorithms, applications to lane markings tracking, road sign detection and recognition, road shape fitting and road surface 3D reconstruction are presented. [Copyright &y& Elsevier]

Abstract: The increasing growth of data on protein–protein interaction (PPI) networks has boosted research on their comparative analysis. In particular, recent studies proposed models and algorithms for performing network alignment, that is, the comparison of networks across species for discovering conserved functional complexes. In this paper, we present an algorithm for dividing PPI networks, prior to their alignment, into small sub-graphs that are likely to cover conserved complexes. This allows one to perform network alignment in a modular fashion, by acting on pairs of resulting small sub-graphs from different species. The proposed dividing algorithm combines a graph-theoretical property (articulation) with a biological one (orthology). Extensive experiments on various PPI networks are conducted in order to assess how well the sub-graphs generated by this dividing algorithm cover protein functional complexes and whether the proposed pre-processing step can be used for enhancing the performance of network alignment algorithms. Source code of the dividing algorithm is available upon request for academic use. [Copyright &y& Elsevier]

Abstract: Boosting has become one of the state-of-the-art techniques in many supervised learning and semi-supervised learning applications. In this paper, we develop a novel boosting algorithm, MTBoost, for multi-task learning problem. Many previous multi-task learning algorithms can only solve the problem in low or moderate dimensional space. However, the MTBoost algorithm is capable of working for very high dimensional data such as in text mining where the feature number is beyond several 10,000. The experimental results illustrate that the MTBoost algorithm provides significantly better classification performance than supervised single task learning algorithms. Moreover, MTBoost outperforms some other typical multi-task learning methods. [ABSTRACT FROM AUTHOR]

Abstract: In this paper, an algorithm for rectifying heterogeneous and uncalibrated pairs of stereo images is presented. In particular, a pair of images is captured by using a combination of static and dynamic cameras at unequal zoom, thus having different focal lengths and/or image resolutions. The rectification of such pairs of images is made in two steps. In the first step, image shrinking based on focal ratios is performed for compensating the effect of unequal zoom levels followed by a zero padding on the smaller image for making the images of equal size. In the second step, rectification transformations are calculated by solving a nonlinear constrained optimization problem for a given set of pairs of corresponding points (SIFT descriptors) between stereo images. Experiments are performed to evaluate the performance of the proposed method and assess the improvements of the proposed method over direct rectification. [Copyright &y& Elsevier]

Abstract: This paper proposes a novel algorithm framework to solve the eye/eyes tracking problem. Two unified deformable templates are introduced for single eye tracking and double eyes tracking, respectively. Each deformable template can describe both open and closed eye states. Based on the templates the particle filtering framework is applied. A dynamic model considering both the geometrical transformations and state transitions of eye/eyes is given. Furthermore, a measurement model for contour tracking is also modified to adapt for eye/eyes tracking. The experimental results show that our approach can not only track the locations of eye/eyes accurately, but also obtain the eye contour parameters simultaneously. [Copyright &y& Elsevier]

Abstract: In this paper, new adaptive learning algorithms and correspondent networks are presented in order to extract optimal features from a sequence of multidimensional Gaussian data. For this purpose, novel adaptive algorithms for the estimation of the square root of the inverse covariance matrix are introduced and applied for Gaussian optimal feature extraction. New adaptive algorithms are drawn by optimization of an appropriate cost function that is presented for the first time. Based on the proposed adaptive algorithms, related networks are implemented in order to extract optimal features from a sequence of multidimensional Gaussian data. Adaptive nature of the new feature extraction methods makes them appropriate for on-line pattern recognition applications. Experimental results using multidimensional Gaussian data demonstrated the effectiveness of the new adaptive feature extraction methods. [Copyright &y& Elsevier]

Abstract: In this paper we describe the implementation of a complete ANN training procedure using the block mode back-propagation learning algorithm for sequential patterns – such as the observation feature vectors of a speech recognition system – exploiting the high performance SIMD architecture of GPU using CUDA and its C-like language interface. We also compare the speed-up obtained implementing the training procedure only taking advantage of the multi-thread capabilities of multi-core processors. In our implementation we take into account all the peculiar aspects of training large scale sequential patterns, in particular, the re-segmentation of the training sentences, the block size for the feed-forward and for the back-propagation steps, and the transfer of huge amount of data from host memory to the GPU card. Our approach has been tested by training acoustic models for large vocabulary speech recognition tasks, showing a six times reduction of the time required to train real-world large size networks with respect to an already optimized implementation using the Intel MKL libraries. Thanks to these optimizations and to the support of the GPU, the training time for language having a huge set of training sentences (about one million for Italian) can be reduced from approximately a month to 5days. [Copyright &y& Elsevier]

Abstract: It is desirable to find unusual data objects by Ramaswamy et al.’s distance-based outlier definition, because only a metric distance function between two objects is required. This definition does not need any neighborhood distance threshold required by many existing algorithms based on the definition of Knorr and Ng. Bay and Schwabacher proposed an efficient algorithm ORCA, which can give near linear time performance, for this task. To further reduce the running time, we propose in this paper two algorithms RC and RS using the following two techniques, respectively: (i) faster cutoff update, and (ii) space utilization after pruning. We tested RC, RS, and RCS (a hybrid approach combining both RC and RS) on several large and high-dimensional real data sets with millions of objects. The experiments show that the speed of RCS is as fast as 1.4–2.3 times that of ORCA, and the improvement of RCS is relatively insensitive to the increase in the data size. [Copyright &y& Elsevier]

Abstract: Automatically detecting text-lines from document images has been long studied. However, most researchers today are focusing on boosting the detection rate instead of noise removal. In this paper, we propose a semi-supervised learning framework that targets to segment Manhattan-layout documents with significant levels of noise. The algorithm consists of three steps: first, an initial segmentation process uses the seed filling algorithm; second, an iterative grouping process uses the projection profiles to estimate the vertical border of page contents; third, an inside page-content noise removal uses the online training and classification. We test our algorithm using two databases. The first is the University of Washington (UW)-III database with 1,600 images of different input qualities that has been widely used by the Document Analysis Research (DAR) communities to measure segmentation algorithm performance. The second is the NILE database created by sampling from 320 journals pages of east Asian, east European and middle Eastern languages. The result shows that our framework achieves competitive performance in terms of both page frame level segmentation and text-line level segmentation, and is particularly strong at filtering noise. It also shows that our algorithm is more adaptive to language variations. [Copyright &y& Elsevier]

Abstract: This paper proposes an active contour framework for spectrogram track detection. A potential energy is proposed which results in feature extraction at a signal-to-noise ratio (SNR) of 0.5dB. We show, through complexity analysis, that this is achievable in real-time. [Copyright &y& Elsevier]

Abstract: Light Detection And Ranging (LIDAR) is an important modality in terrain and land surveying for many environmental, engineering and civil applications. This paper presents the framework for a recently developed unsupervised classification algorithm called Skewness Balancing for object and ground point separation in airborne LIDAR data. The main advantages of the algorithm are threshold-freedom and independence from LIDAR data format and resolution, while preserving object and terrain details. The framework for Skewness Balancing has been built in this contribution with a prediction model in which unknown LIDAR tiles can be categorised as “hilly” or “moderate” terrains. Accuracy assessment of the model is carried out using cross-validation with an overall accuracy of 95%. An extension to the algorithm is developed to address the overclassification issue for hilly terrain. For moderate terrain, the results show that from the classified tiles detached objects (buildings and vegetation) and attached objects (bridges and motorway junctions) are separated from bare earth (ground, roads and yards) which makes Skewness Balancing ideal to be integrated into geographic information system (GIS) software packages. [Copyright &y& Elsevier]

Abstract: Bio-based human computer interface (HCI) has attracted more and more attention of researchers all over the world in recent years. The paper is concerned with eye movement detection algorithm and the EOG-based human computer interface. The linear predictive coding cepstrum (LPCC) coefficients of EOG pulses are extracted as feature vectors, which are used for eye movement pattern matching. Dynamic time warping (DTW) is adopted to solve the discrepancy in EOG pulses duration among different trials. Furthermore, spectral entropy algorithm is used to detect the endpoints of EOG pulses, which can improve the robustness and increase the recognition rate in noisy background. Experimental and simulation results based on real-life EOG signals show that the proposed algorithm has stable performance and can be used for online controlling and communication in EOG based HCI system. [Copyright &y& Elsevier]

Abstract: Despite of its great success, two key problems are still unresolved for AdaBoost algorithms: how to select the most discriminative weak learners and how to optimally combine them. In this paper, a new AdaBoost algorithm is proposed to make improvement in the two aspects. First, we select the most discriminative weak learners by minimizing a novel distance related criterion, i.e., error-degree-weighted training error metric (ETEM) together with generalization capability metric (GCM), rather than training error rate only. Second, after getting the coefficients that are set empirically, we combine the weak learners optimally by tuning the coefficients using kernel-based perceptron. Experiments with synthetic and real scene data sets show our algorithm outperforms conventional AdaBoost. [Copyright &y& Elsevier]

Abstract: Face and iris localization is one of the most active research areas in image understanding for new applications in security and theft prevention, as well as in the development of human–machine interfaces. In the past, several methods for real-time face localization have been developed using face anthropometric templates which include face features such as eyes, eyebrows, nose and mouth. It has been shown that accuracy in face and iris localization is crucial to face recognition algorithms. An error of a few pixels in face or iris localization will produce significant reduction in face recognition rates. In this paper, we present a new method based on particle swarm optimization (PSO) to generate templates for frontal face localization in real time. The PSO templates were tested for face localization on the Yale B Face Database and compared to other methods based on anthropometric templates and Adaboost. Additionally, the PSO templates were compared in iris localization to a method using combined binary edge and intensity information in two subsets of the AR face database, and to a method based on SVM classifiers in a subset of the FERET database. Results show that the PSO templates exhibit better spatial selectivity for frontal faces resulting in a better performance in face localization and face size estimation. Correct face localization reached a rate of 97.4% on Yale B which was higher than 96.2% obtained with the anthropometric templates and much better than 60.5% obtained with the Adaboost face detection method. On the AR face subsets, different disparity errors were considered and for the smallest error, a 100% correct detection was reached in the AR-63 subset and 99.7% was obtained in the AR-564 subset. On the FERET subset a detection rate of 96.6% was achieved using the same criteria. In contrast to the Adaboost method, PSO templates were able to localize faces on high-contrast or poorly illuminated environments. Additionally, in comparison with the anthropometric templates, the PSO templates have fewer pixels, resulting in a 40% reduction in processing time thus making them more appropriate for real-time applications. [Copyright &y& Elsevier]