Papers by Fakhita Regragui
HAL (Le Centre pour la Communication Scientifique Directe), Sep 5, 2011
-Nous proposons dans ce papier la combinaison des caractéristiques statistiques locales et global... more -Nous proposons dans ce papier la combinaison des caractéristiques statistiques locales et globales en segmentation par contour actif. L'évaluation de la performance de l'approche proposée sur différents types d'images, présentant des attributs hétérogènes avec une initialisation inadéquate du contour actif ou avec la présence du bruit, a donné des résultats satisfaisants, même quand les méthodes conventionnelles ne parviennent pas à segmenter correctement l'objet d'intérêt.
Materials evaluation, Feb 1, 2015
ABSTRACT In the field of nondestructive testing of welded components, the most important stages o... more ABSTRACT In the field of nondestructive testing of welded components, the most important stages of automatic inspection systems concern the detection and classification of weld discontinuities. Limitations to correlating the heterogeneity and the discontinuity are imposed by the nature of the discontinuity: morphology, position, orientation, size, and so forth. Commonly seen weld discontinuities include cracks, linear inclusions, lack of penetration, and porosities. In their previous article, the authors were interested in discontinuity detection. In this paper, the authors attempted to provide a complete analysis for flaw classification taking into account the problem of data set unbalance using a synthetic minority oversampling technique. In addition, a cross-validated linear support vector machine-based recursive feature elimination algorithm was developed to perform feature selection, allowing better generalization performance.
A novel video watermarking system operating in the three-dimensional wavelet transform is here pr... more A novel video watermarking system operating in the three-dimensional wavelet transform is here presented. S pecifically the video sequence is partitioned into spatiotemporal units and the single shots are projected onto the 3D wavelet domain. first a gray-scale watermark image is decomposed into a series of bitplanes that are preprocessed with a random location matrix. after that the preprocessed bitplanes are adaptively spread spectrum and added in 3D wavelet coefficients of the video shot. Our video watermarking algorithm is robust against the attacks of frame dropping, averaging and swapping. Furthermore, it allows blind retrieval of embedded watermark which does not need the original video and the watermark is perceptually invisible. The algorithm design, evaluation, and experimentation of the proposed scheme are described in this paper.
Computational and Mathematical Methods in Medicine, 2012
Functional magnetic resonance imaging (fMRI) exploits blood-oxygen-level-dependent (BOLD) contras... more Functional magnetic resonance imaging (fMRI) exploits blood-oxygen-level-dependent (BOLD) contrasts to map neural activity associated with a variety of brain functions including sensory processing, motor control, and cognitive and emotional functions. The general linear model (GLM) approach is used to reveal task-related brain areas by searching for linear correlations between the fMRI time course and a reference model. One of the limitations of the GLM approach is the assumption that the covariance across neighbouring voxels is not informative about the cognitive function under examination. Multivoxel pattern analysis (MVPA) represents a promising technique that is currently exploited to investigate the information contained in distributed patterns of neural activity to infer the functional role of brain areas and networks. MVPA is considered as a supervised classification problem where a classifier attempts to capture the relationships between spatial pattern of fMRI activity and experimental conditions. In this paper , we review MVPA and describe the mathematical basis of the classification algorithms used for decoding fMRI signals, such as support vector machines (SVMs). In addition, we describe the workflow of processing steps required for MVPA such as feature selection, dimensionality reduction, cross-validation, and classifier performance estimation based on receiver operating characteristic (ROC) curves.
Results in engineering, Sep 1, 2022
HAL (Le Centre pour la Communication Scientifique Directe), Sep 11, 2011
This paper proposes a novel approach that allows regionbased active contour energy to be re-expre... more This paper proposes a novel approach that allows regionbased active contour energy to be re-expressed in a local and global manner. The basic idea of this technique consists in extracting locally image statistics from the heterogeneous region (foreground or background) and globally from the other region at each point along the curve. By exploiting benefits of both local-based and global-based statistics, this technique proves to be robust against heterogeneity and noise and shows low sensitivity to curve initialization. Experimental results for synthetic and real images reveal significant improvement compared to conventional methods.
X-ray radiography is one of the most used techniques in the non destructive testing(NDT). It allo... more X-ray radiography is one of the most used techniques in the non destructive testing(NDT). It allows the detection of weld defects the most dangerous for the weld's integrity. Because X-ray images of welds are noisy and low contrasted, it is difficult to detect weld defects inside. The goal of this paper is to segment the defects in X-ray images. However,
arXiv (Cornell University), Nov 2, 2009
A novel video watermarking system operating in the three-dimensional wavelet transform is here pr... more A novel video watermarking system operating in the three-dimensional wavelet transform is here presented. Specifically the video sequence is partitioned into spatio-temporal units and the single shots are projected onto the 3D wavelet domain. First a gray-scale watermark image is decomposed into a series of bitplanes that are preprocessed with a random location matrix. After that the preprocessed bitplanes are adaptively spread spectrum and added in 3D wavelet coefficients of the video shot. Our video watermarking algorithm is robust against the attacks of frame dropping, averaging and swapping. Furthermore, it allows blind retrieval of embedded watermark which does not need the original video and the watermark is perceptually invisible. The algorithm design, evaluation, and experimentation of the proposed scheme are described in this paper.
HAL (Le Centre pour la Communication Scientifique Directe), Jun 15, 2015
Dans ce papier, nous proposons une méthode de reconnaissance 3D des gestes pour l'interaction hom... more Dans ce papier, nous proposons une méthode de reconnaissance 3D des gestes pour l'interaction homme robot (HRI) basée sur l'information de profondeur fournie par la Kinect. Le suivi du corps est réalisé avec l'algorithme Skeleton fourni par le Kinect SDK. L'idée de ce travail est de calculer les angles des articulations de la partie supérieure du corps durant l'exécution du geste. Les variations de ces angles seront les entrées des Modèles de Markov Cachés afin de reconnaître les gestes dynamiques. Les résultats montrent que notre méthode est très robuste ; elle nécessite peu de prétraitements et n'est pas influencée par les conditions de l'environnement comme les changements d'éclairage et la complexité de la scène.
DOAJ (DOAJ: Directory of Open Access Journals), Jun 1, 2007
2016 Sixth International Conference on Image Processing Theory, Tools and Applications (IPTA), 2016
Gesture recognition is one of the important tasks for human System Interaction (HRI). This paper ... more Gesture recognition is one of the important tasks for human System Interaction (HRI). This paper describes a novel approach intended to recognize 3D dynamic composed gestures by combining Dynamic Time Warping (DTW) with an Adaptive Sliding Window which the name Adaptive Dynamic Time Warping (ADTW). We use the skeleton algorithm provided by the Kinect SDK to track the upper part of body and extract joints angles based on depth information. Each gesture is represented by the combination of angles variations and stored described as a vector. A composed gesture is a sequence of two simple gestures or more performed successively in time. We chose five simple gestures : come, recede, point to the right, point to the left and stop. For each simple gesture, we chose a reference sequence that perfectly represents it. In order to recognize all gesture of the composed gesture in the right order, we combine (DTW) with an Adaptive Sliding Window. In one hand, we use an adaptive window to browse through the sequence of the composed gesture by feeding it to each time with new data. In other hand, we use DTW to compare between the reference gestures and the the sequences defined by the adaptive window. In fact, by comparing each two sequences, DTW computes the euclidean distance between them. Finally, the reference gesture which gives the lower distance is considered as the source class of the tested gesture.
2016 11th International Conference on Intelligent Systems: Theories and Applications (SITA), 2016
Brain-machine interface (BMI) is a system that allows a person to control a device such as a robo... more Brain-machine interface (BMI) is a system that allows a person to control a device such as a robot arm using only his or her brain activity. This work is aimed at discriminating between left and right imagined hand movements using a Support Vector Machine (SVM) classifier. The main focus here is to search for the best features that efficiently describe the electroencephalogram (EEG) data during such imagined gestures. The EEG dataset used in this research was recorded using channels F3 and F4 from the Emotiv EPOC neural headset. Feature extraction was performed by processing the EEG data using two methods namely the continuous Wavelet Transform (CWT) combined with the Principal Component Analysis (PCA). The features were fed through a Linear and RBF Kernel SVM classifier. The Experimental results showed high performance achieving an average accuracy across all the subjects of 92.75% with a RBF kernel SVM classifier compared to 81.12% accuracy obtained with Linear SVM classifier.
Applied Mathematical Sciences, 2013
In this paper, we propose a hybrid object tracking approach based on region intensities and the m... more In this paper, we propose a hybrid object tracking approach based on region intensities and the motion vector of the interest points of the target object. After the heterogeneous region is identified (i.e foreground or background), the information extracted from the region intensities is computed through a pre-processing selection technique using local and global statistics. The motion vector of the interest points is obtained by computing the mean displacement of these points between consecutive frames. Therefore, for each received frame, the displacement vector is applied to the initial active contour that will be evolved using the region information. The main advantages of our approach which combines the information gathered from region intensities and motion of the interest points are first, a priori knowledge of the heterogeneous region is not required and second, the initial active contour in each frame is adjusted much closer to the true boundary of the object of interest. Experiments with synthetic and real-world images validate the efficiency of the proposed hybrid approach.
Computers in Biology and Medicine, Apr 1, 2006
A non-linear classifier is proposed to discriminate visual evoked potentials (VEP). It combines t... more A non-linear classifier is proposed to discriminate visual evoked potentials (VEP). It combines two techniques: the zero-tracking method and a multi-layer network. The first method consists of processing the VEP data through an adaptive linear prediction filter aiming at extracting the appropriate feature vector to be fed into the neural network. 105 VEPs collected from 48 healthy people and 57 patients are analysed to test the performances of the proposed classifier. The results obtained with a back-propagation network revealed a total success rate equal to 89%. It is also found more accurate than the latency method used in hospitals.
Le Centre pour la Communication Scientifique Directe - HAL - Inria, Jun 15, 2015
Dans ce papier, nous proposons une méthode de reconnaissance 3D des gestes pour l'interaction hom... more Dans ce papier, nous proposons une méthode de reconnaissance 3D des gestes pour l'interaction homme robot (HRI) basée sur l'information de profondeur fournie par la Kinect. Le suivi du corps est réalisé avec l'algorithme Skeleton fourni par le Kinect SDK. L'idée de ce travail est de calculer les angles des articulations de la partie supérieure du corps durant l'exécution du geste. Les variations de ces angles seront les entrées des Modèles de Markov Cachés afin de reconnaître les gestes dynamiques. Les résultats montrent que notre méthode est très robuste ; elle nécessite peu de prétraitements et n'est pas influencée par les conditions de l'environnement comme les changements d'éclairage et la complexité de la scène.
Microcalcifications can be a very important sign of breast cancer. As their detection is very cru... more Microcalcifications can be a very important sign of breast cancer. As their detection is very crucial to further investigation, automatic detection in mammograms can help practitioners to locate missed abnormalities. The aim of this work is to propose a simple method based on fuzzy clustering to efficiently segment microcalcifications. This method which is derived from two existing methods, automatically determines the number of classes in each image and then isolates potential microcalcifications. Compared to previous methods, the proposed method was tested on 7 Regions of Interest and demonstrated higher performance reaching up to 0.93 in terms of F1-Score and an overall best performace.
Dans ce papier, nous proposons une methode de reconnaissance 3D des gestes pour l’interaction hom... more Dans ce papier, nous proposons une methode de reconnaissance 3D des gestes pour l’interaction homme robot (HRI) basee sur l’information de profondeur fournie par la Kinect. Le suivi du corps est realise avec l’algorithme Skeleton fourni par le Kinect SDK. L’idee de ce travail est de calculer les angles des articulations de la partie supe- rieure du corps durant l’execution du geste. Les variations de ces angles seront les entrees des Modeles de Markov Caches afin de reconnaitre les gestes dynamiques. Les re- sultats montrent que notre methode est tres robuste ; elle necessite peu de pretraitements et n’est pas influencee par les conditions de l’environnement comme les changements d’eclairage et la complexite de la scene.
Single visual evoked potentials (VEP) are very weak and noisy signals. For this reason, powerful ... more Single visual evoked potentials (VEP) are very weak and noisy signals. For this reason, powerful extraction tools are needed to improve their clinical use. In this paper, we present two methods for filtering VEP: a linear one, based on the adaptive noise canceller scheme where the input signals represent successive recorded VEP, and a new non-linear method which exploits the neural network property to give a universal approximation of any non-linear function as complex as the VEP signal. We test the effectiveness of these methods and compare their performance to conventional averaging traditionally used in hospitals. As
Uploads
Papers by Fakhita Regragui