By Topic

Image Analysis and Processing, 2007. ICIAP 2007. 14th International Conference on

Date 10-14 Sept. 2007

Filter Results

Displaying Results 1 - 25 of 139
  • 14th International Conference on Image Analysis and Processing-Title

    Page(s): i - iii
    Save to Project icon | Request Permissions | PDF file iconPDF (46 KB)  
    Freely Available from IEEE
  • 14th International Conference on Image Analysis and Processing-Copyright

    Page(s): iv
    Save to Project icon | Request Permissions | PDF file iconPDF (52 KB)  
    Freely Available from IEEE
  • 14th International Conference on Image Analysis and Processing - TOC

    Page(s): v - xiii
    Save to Project icon | Request Permissions | PDF file iconPDF (62 KB)  
    Freely Available from IEEE
  • Foreword

    Page(s): xiv - xv
    Save to Project icon | Request Permissions | PDF file iconPDF (43 KB)  
    Freely Available from IEEE
  • Organizers and Sponsors

    Page(s): xvi
    Save to Project icon | Request Permissions | PDF file iconPDF (148 KB)  
    Freely Available from IEEE
  • Committees

    Page(s): xvii
    Save to Project icon | Request Permissions | PDF file iconPDF (59 KB)  
    Freely Available from IEEE
  • Collarette Area Localization and Asymmetrical Support Vector Machines for Efficient Iris Recognition

    Page(s): 3 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (470 KB) |  | HTML iconHTML  

    This paper presents an efficient iris recognition technique based on the zigzag collarette area localization and asymmetrical support vector machine. The deterministic feature sequence extracted from the iris images using the ID log-Gabor filters is applied to train the support vector machine (SVM). We use the multi- objective genetic algorithm (MOGA) to optimize the features and also to increase the overall recognition accuracy. The traditional SVM is modified to an asymmetrical SVM to treat the cases of the False Accept and the False Reject differently and also to handle the unbalanced data of a specific class with respect to the other classes. The proposed technique is computationally effective with a recognition rate of 97.70% on the ICE (Iris Challenge Evaluation) iris dataset. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Colour and Geometric based Model for Lip Localisation: Application for Lip-reading System

    Page(s): 9 - 14
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (860 KB) |  | HTML iconHTML  

    Motivated by humans' ability to lipread, the visual component is considered to yield information in the speech recognition system. The lip-reading is the perception of the speech purely based on observing the talker lip movements. The major difficulty of the lip- reading system is the extraction of the visual speech descriptors. In fact, to ensure this task it is necessary to carry out an automatic localization and tracking of the labial gestures. We present in this paper a new automatic approach for lip and point of interest localization on a speaker's face based both on the color information of mouth and a geometric model of lips. This hybrid solution makes our method more tolerant to noise and artifacts in the image. Experiments revealed that our lip POI localization approach for lip-reading purpose is promising. The presented results show that our system recognizes 94.64 % of French visemes. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Dynamic Score Selection for Fusion of Multiple Biometric Matchers

    Page(s): 15 - 22
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (189 KB) |  | HTML iconHTML  

    A biometric system for user authentication produces a matching score representing the degree of similarity of the input biometry with the set of templates for that user. If the score is greater than a prefixed threshold, then the user is accepted, otherwise the user is rejected. Typically the performance are evaluated in terms of the receiver operating characteristic (ROC) curve, and the equal error rate (EER). In order to increase the reliability of authentication through biometrics, the combination of different biometric systems is currently investigated by researchers. While a number of "fusion" algorithms have been proposed in the literature, in this paper we propose a theoretical analysis of a novel approach based on the "dynamic selection" of matching scores. Such a selector aims at choosing, for each user to be authenticated, just one of the scores produced by the different biometric systems available. We show that the "dynamic selection" of matching scores can provide a better ROC than those of individual biometric systems. Reported results on the FVC2004 dataset confirm the theoretical analysis, and show that the proposed "dynamic selection" approach is more effective when low quality scores are used. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Integrating Boundary Information in Pairwise Segmentation

    Page(s): 23 - 28
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (580 KB) |  | HTML iconHTML  

    Proximity-based, or pairwise, data clustering techniques are gaining increasing popularity due to their versatility and their ability to easily integrate information of different nature. Despite this, most applications to image segmentation incorporate only region-based information, mainly color and texture similarity. In this paper we propose a general approach for integrating boundary information in a pairwise segmentation framework. To this end we propose a measure of distance between pair of pixels that integrates the value of an edge-response function along a path joining the two pixels. Experiments performed using the dominant sets framework show that the proposed approach is competitive with the state of the art pairwise segmentation algorithms even while using boundary information only. Furthermore, we show that the approach can effectively be used when adopting an out of sample approach to pairwise segmentation. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Kernelised Relaxation Labelling using Fokker-Planck Diffusion

    Page(s): 29 - 34
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (349 KB) |  | HTML iconHTML  

    This paper shows how the relaxation labelling problem can be formulated as a diffusion process on a support graph using the Fokker-Planck equation. We abstract the labelling problem using a support graph with each graph node representing a possible object-label assignment and the edge weights representing label compatibilities. Initial object-label probabilities are updated using a relaxation-like process. The update equation is the solution of the Fokker-Planck equation, and is governed by an infinitesimal generator matrix computed from the edge-weights of the support graph. Iterative updating of the label probabilities can be effected using the eigenvalues and eigenvectors of the generator matrix. We illustrate the newly developed relaxation process for the applications of data classification. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Spectral Generative Models for Graphs

    Page(s): 35 - 42
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (632 KB) |  | HTML iconHTML  

    Generative models are well known in the domain of statistical pattern recognition. Typically, they describe the probability distribution of patterns in a vector space. The individual patterns are defined by vectors and so the individual features of the pattern are well defined. In contrast, very little work has been done with generative models of graphs because graphs do not have a straightforward vectorial representation. Because of this, simple statistical quantities such as mean and variance are difficult to define for a group of graphs. While we can define statistical quantities of individual edges, it is not so straightforward to define how sets of edges in graphs are related. In this paper we examine the problem of creating generative distributions over sets of graphs. We use the spectral representation of the graphs to construct a dual vector space for the graphs. The spectral decomposition of a graph can be used to extract information about the relationship of edges and parts in a graph. Distributions are then defined on the vector spaces and used to generate new samples. Finally, these points must be used to reconstruct the sampled graph. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Facing Imbalanced Classes through Aggregation of Classifiers

    Page(s): 43 - 48
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (709 KB) |  | HTML iconHTML  

    Two class classification problems in real world are often characterized by imbalanced classes. This is a serious issue since a classifier trained on such a data distribution typically exhibits a prediction accuracy highly skewed towards the majority class. To improve the quality of the classifier, many approaches have been proposed till now for building artificially balanced training sets. Such methods are mainly based on undersampling the majority class and/or oversampling the minority class. However, both approaches can produce overfitting or underfitting problems for the trained classifier. In this paper we present a method for building a multiple classifier system in which each constituting classifier is trained on a subset of the majority class and on the whole minority class. The approach has been tested on the detection of microcalcifications on digital mammograms. The results obtained confirm the effectiveness of the method. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Learning Repetitive Patterns for Classifying Non-Rigidly Deforming Texture Surfaces

    Page(s): 49 - 54
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (3448 KB) |  | HTML iconHTML  

    In this paper, we address the relatively unexplored problem of classifying texture surfaces undergoing significant levels of non-rigid deformation. State-of-the-art texture classification methods have demonstrated to be very effective for classifying fronto-parallel texture fields. Recently, affine-invariant descriptors have been proposed as an effective way to model local perspective distortion in textures. However, if the effects of local surface curvature distortion are large, affine-invariant descriptors become unreliable. Our contribution in this paper is twofold. First, we propose a method for learning representative basic elements of non-fronto-parallel texture fields undergoing non-rigid deformations. Secondly, we demonstrate the effectiveness of our texture learning method for the classification of non-rigid deforming texture surfaces. We test our method on a set of images obtained from man-made texture surfaces. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Motion Estimation via Belief Propagation

    Page(s): 55 - 60
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (406 KB) |  | HTML iconHTML  

    We present a probabilistic model for motion estimation in which motion characteristics are inferred on the basis of a finite mixture of motion models. The model is graphically represented in the form of a pairwise Markov random field network upon which a Loopy belief propagation algorithm is exploited to perform inference. Experiments on different video clips are presented and discussed. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Performance Evaluation of Scale-Interpolated Hessian-Laplace and Haar Descriptors for Feature Matching

    Page(s): 61 - 66
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (213 KB) |  | HTML iconHTML  

    This paper studies the performance of various scale- invariant detectors in the context of feature matching. In particular, we propose an implementation of the Hessian-Laplace operator that we called scale-interpolated Hessian-Laplace. This research also proposes to use Haar descriptors which are derived from the Haar wavelet transform. It offers the advantage of being computationally inexpensive and smaller in size when compared to other descriptors. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Sparseness Achievement in Hidden Markov Models

    Page(s): 67 - 72
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (282 KB) |  | HTML iconHTML  

    In this paper, a novel learning algorithm for Hidden Markov Models (HMMs) has been devised. The key issue is the achievement of a sparse model, i.e., a model in which all irrelevant parameters are set exactly to zero. Alternatively to standard maximum likelihood estimation (Baum Welch training), in the proposed approach the parameters estimation problem is cast into a Bayesian framework, with the introduction of a negative Dirichlet prior, which strongly encourages sparseness of the model. A modified Expectation Maximization algorithm has been devised, able to determine a MAP (maximum a posteriori probability) estimate of HMM parameters in this Bayesian formulation. Theoretical considerations and experimental comparative evaluations on a 2D shape classification task contribute to validate the proposed technique. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Using Bayesian Network for combining classifiers

    Page(s): 73 - 80
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (222 KB) |  | HTML iconHTML  

    In the framework of multiple classifier systems, we suggest to reformulate the classifier combination problem as a pattern recognition one. Following this approach, each input pattern is associated to a feature vector composed by the output of the classifiers to be combined. A Bayesian Network is used to automatically infer the probability distribution for each class and eventually to perform the final classification. We propose to use Bayesian Networks because they not only provide a basis for efficient probabilistic inference, but also a natural and compact way to encode exponentially sized joint probability distributions. Two systems adopting an ensemble of Back-Propagation neural network and an ensemble of Learning Vector Quantization neural network, respectively, have been tested on the Image database from the UCI repository. The performance of the proposed systems have been compared with those exhibited by multi-expert systems adopting the same ensembles, but the Majority Vote, the Weighted Majority vote and the Borda Count for combining them. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A Method of Clustering Combination Applied to Satellite Image Analysis

    Page(s): 81 - 86
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (140 KB) |  | HTML iconHTML  

    An algorithm for combining results of different clusterings is presented in this paper, the objective of which is to find groups of patterns which are common to all clusterings. The idea of the proposed combination is to group those samples which are in the same cluster in most cases. We formulate the combination as the resolution of a linear set of equations with binary constraints. The advantage of such a formulation is to provide an objective function for the combination. To optimize the objective function we propose an original unsupervised algorithm. Furthermore, we propose an extension adapted in case of a huge volume of data. The combination of clusterings is performed on the results of different clustering algorithms applied to SPOT5 satellite images and shows the effectiveness of the proposed method. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Ball Position and Motion Reconstruction from Blur in a Single Perspective Image

    Page(s): 87 - 92
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (3001 KB) |  | HTML iconHTML  

    We consider the problem of localizing a moving ball from a single calibrated perspective image; after showing that ordinary algorithms fail in analyzing motion blurred scenes, we describe a theoretically-sound model for the blurred image of a ball. Then, we present an algorithm capable of recovering both the ball 3D position and its velocity. The algorithm is experimentally validated both on real and synthetic images. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Calibration and Image Generation of Mobile Projector-Camera Systems

    Page(s): 93 - 98
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (478 KB) |  | HTML iconHTML  

    The projector-camera system has recently been studied extensively as one of new information presenting systems. For generating screen images properly, it is important to calibrate projector-camera systems accurately. The existing methods for calibrating projector-camera systems are based on 4 markers on the screen and 4 light projections from projectors, and thus require at least 8 basis points totally in images. However, it is not easy to track 8 or more basis points reliably in images, if the projector camera system moves arbitrarily. Thus, we in this paper propose a method for generating screen images properly from less basis points in camera images. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Hybrid Stereo Sensor with Omnidirectional Vision Capabilities: Overview and Calibration Procedures

    Page(s): 99 - 104
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (238 KB) |  | HTML iconHTML  

    In this paper, we present a compact hybrid video sensor that combines perspective and omnidirectional vision to achieve a 360deg field of view, as well as high-resolution images. Those characteristics, in association with 3D metric reconstruction capabilities, are suitable for vision tasks such as surveillance and obstacle detection for autonomous robot navigation. We describe the sensor calibration procedure, with particular regard to mirror-to-camera positioning. We also present some results obtained in testing the accuracy of 3D reconstruction, which have confirmed the correctness of the calibration. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Image Spam Filtering Using Visual Information

    Page(s): 105 - 110
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (280 KB) |  | HTML iconHTML  

    We address the problem of recognizing the so-called image spam, which consists in embedding the spam message into attached images to defeat techniques based on the analysis of e-mails' body text, and in using content obscuring techniques to defeat OCR tools. We propose an approach to recognize image spam based on detecting the presence of content obscuring techniques, and describe a possible implementation based on two low-level image features aimed at detecting obscuring techniques whose consequence is to compromise the OCR effectiveness resulting in character breaking or merging, or in the presence of noise interfering with characters in the binarized image. A preliminary experimental investigation of this approach is reported on a personal data set of spam images. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Localization of ahead vehicles with on-board stereo cameras

    Page(s): 111 - 116
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (515 KB) |  | HTML iconHTML  

    This paper introduces a vision based algorithm that detects and localizes ahead vehicles elaborating images taken by a stereo camera installed on an intelligent vehicle. The algorithm is based on the analysis of stereo images, estimating the ground plane by least square fitting of disparity data, and segmenting the obstacles by a rule based split/merge strategy. Quantitative experiments on complex real world sequences validate the approach. The method is demonstrated to operate in real-time. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Rectification of 3D Data Obtained from Moving Range Sensors by using Multiple View Geometry

    Page(s): 117 - 122
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1909 KB) |  | HTML iconHTML  

    For measuring the 3D shape of large objects, scanning by a moving range sensor is one of the most efficient method. However, if we use moving range sensors, the obtained data have some distortions due to the movement of the sensor during the scanning process. In this paper, we propose a method for recovering correct 3D range data from a moving range sensor by using the multiple view geometry. We assume that range sensor radiates laser beams in raster scan order, and they are observed from a static camera. We first show that we can deal with range data as 3D space-time images, and show that the extended multiple view geometry can be used for representing the relationship between the 3D space-time of camera image and the 3D space-time of range data. We next show that the multiple view geometry under extended projections can be used for rectifying 3D data obtained by the moving range sensor. The method is implemented and tested in synthetic images and range data. The stability of recovered 3D shape is also evaluated. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.