Browsing by Author "Setitra, Insaf "

Now showing 1 - 17 of 17
  • Thumbnail Image
    Item
    A framework for Object Classification in Fareld Videos
    (Springer, 2014-12) Setitra, Insaf; Larabi, Slimane
    Object classification in videos is an important step in many applications such as abnormal event detection in video surveillance, traf- fic analysis is urban scenes and behavior control in crowded locations. In this work, propose a framework for moving object classification in farfield videos. Much works have been dedicated to accomplish this task. We overview existing works and combine several techniques to implement a real time object classifier with offline training phase. We follow three main steps to classify objects in steady background videos : background subtraction, object tracking and classification. We measure accuracy of our classifier by experiments done using the PETS 2009 dataset.
  • No Thumbnail Available
    Item
    A framework for object classification in farfield videos
    (CERIST, 2014-10-26) Setitra, Insaf; Larabi, Slimane
    Object classification in videos is an important step in many applications such as abnormal event detection in video surveillance, traffic analysis is urban scenes and behavior control in crowded locations. In this work, propose a framework for moving object classification in farfield videos. Much works have been dedicated to accomplish this task. We overview existing works and combine several techniques to implement a real time object classifier with offline training phase. We follow three main steps to classify objects in steady background videos : background subtraction, object tracking and classification. We measure accuracy of our classifier by experiments done using the PETS 2009 dataset.
  • Thumbnail Image
    Item
    A simple approach to distinguish between Maghrebi and Persian calligraphy in old manuscripts
    (IAPR, 2014-04-07) Setitra, Insaf; Meziane, Abdelkrim
    Manual annotation of images is usually a mandatory task in many applications where no knowledge about the image is available. In presence of huge number of images, this task becomes very tedious and prone to human errors. In this paper, we contribute in automatic annotation of Arabic old manuscripts by discovering manuscript calligraphy. Arabic manuscripts count a very large number of Persian and Maghrebi writing especially in Noth Africa. Distinguishing between these two calligraphies allow better classifying them and so annotating them. We use background constructing followed by extraction of simple features to classify Arabic manuscript calligraphies using a Quadratic Bayes classifier.
  • No Thumbnail Available
    Item
    A simple approach to distinguish between Maghrebi and Persian calligraphy in old manuscripts
    (CERIST, 2014-04-07) Setitra, Insaf; Meziane, Abdelkrim
    Manual annotation of images is usually a mandatory task in many applications where no knowledge about the image is available. In presence of huge number of images, this task becomes very tedious and prone to human errors. In this paper, we contribute in automatic annotation of Arabic old manuscripts by discovering manuscript calligraphy. Arabic manuscripts count a very large number of Persian and Maghrebi writing especially in Noth Africa. Distinguishing between these two calligraphies allow better classifying them and so annotating them. We use background constructing followed by extraction of simple features to classify Arabic manuscript calligraphies using a Quadratic Bayes classifier.
  • Thumbnail Image
    Item
    A study on discrimination of SIFT feature applied to binary images
    (CERIST, 2015-10-04) Setitra, Insaf; Larabi, Slimane
    Scale Invariant Feature Transform (SIFT) since its first apparition in 2004 has been (and still is) extensively used in computer vision to classify and match objects in RGB and grey level images and videos. However, since the descriptor used in SIFT approach is based on gradient magnitude and orientation, it has always been considered as texture feature and received less interest when treating binary images. In this work we investigate the power of discrimination of SIFT applied to binary images. A theoretical and experimental studies show that SIFT can still describe shapes and can be used to distinguish objects of several classes.
  • No Thumbnail Available
    Item
    A Tracking Approach for Text Line Segmentation in Handwritten Documents
    (Springer / LNCS Series Book, 2017-02-24) Setitra, Insaf
    Tracking of objects in videos consists of giving a label to the same object moving in different frames. This labelling is performed by predicting position of the object given its set of features observed in previous frames. In this work, we apply the same rationale by considering each connected component in the manuscript as a moving object and to track it so that to minimize the distance and angle of of the connected component to its nearest neighbour. The approach was applied to images of ICDAR 2013 handwritten segmentation contest and proved to be robust against text orientation, size and writing script.
  • Thumbnail Image
    Item
    A tracking approach for text line segmentation in handwritten documents
    (Springer, 2017-02-24) Setitra, Insaf; Hadjadj, Zineb; Meziane, Abdelkrim
    Tracking of objects in videos consists of giving a label to the same object moving in different frames. This labeling is performed by predicting position of the object given its set of features observed in previous frames. In this work, we apply the same rationale by considering each connected component in the manuscript as a moving object and to track it so that to minimize the distance and angle of the connected component to its nearest neighbor. The approach was applied to images of ICDAR 2013 handwritten segmentation contest and proved to be robust against text orientation, size and writing script.
  • Thumbnail Image
    Item
    A tracking approach for text line segmentation in handwritten documents
    (CERIST, 2017-02-24) Setitra, Insaf; Hadjadj, Zineb; Meziane, Abdelkrim
    Tracking of objects in videos consists of giving a label to the same object moving in different frames. This labeling is performed by predicting position of the object given its set of features observed in previous frames. In this work, we apply the same rationale by considering each connected component in the manuscript as a moving object and to track it so that to minimize the distance and angle of the connected component to its nearest neighbor. The approach was applied to images of ICDAR 2013 handwritten segmentation contest and proved to be robust against text orientation, size and writing script.
  • No Thumbnail Available
    Item
    Angle Minimization and Graph Analysis for text line segmentation in handwritten documents
    (CERIST, 2018-07-08) Setitra, Insaf; Meziane, Abdelkrim
    We propose in this paper a novel approach for text line segmentation in handwritten documents. The approach is based on angle minimization and graph analysis for text lines extraction. We apply our approach on images of ICDAR 2013 Handwriting Segmentation Contest, and give details about its robustness against skew and text orientation. We compare the approach to relevant text line segmentation state of art methods, apply it to Algerian manuscripts and report relevant results
  • No Thumbnail Available
    Item
    Background subtraction algorithms with post processing A review
    (CERIST, 2014-04-28) Setitra, Insaf; Larabi, Slimane
    Due to its several algorithms with their fast implementations, background subtraction becomes a very important step in many computer vision and video surveillance systems which assume static cameras. Literature counts a large number of robust background subtraction algorithms which try each to outperform the others in a quantitative and qualitative manner. This competition can sometimes confuse the user of this kind of process and make the choice of one of them difficult. To overcome this issue we review, in what follows, the background subtraction process by defining it and exploring most used algorithms of background subtraction. We then expose some post processing techniques used to remove superfluous content derived from background subtraction.
  • Thumbnail Image
    Item
    Classification automatique des images histologiques du cancer du sein par réseaux de neurones convolutifs (RNC)
    (Publication en ligne, 2018-08-01) Setitra, Insaf; Meziane, Abdelkrim; Mayouf, Mouna Sabrine; Hamrioui, Amel
    Après le cancer de la peau, le cancer du sein est le deuxième type de cancer le plus commun chez la femme à l’échelle mondiale. Ce dernier enregistre un taux de mortalité assez élevé comparé aux autres types de cancer. (Spanhol, Oliveira et al. 2016). Le diagnostic des tumeurs du sein pour différencier les cellules bénignes des malignes établi par le pathologiste est le fruit d’un processus minutieux, fastidieux, long et sujet à plusieurs erreurs et divergence d’avis. Afin d’essayer de palier à ces inconvénients, un vif intérêt s’est porté sur l’automatisation du processus du diagnostic. Dans ce travail, nous reportons les différentes méthodes utilisées jusque-là par la communauté scientifique et nous exposons notre méthode basée sur la classification par réseaux de neurones convolutifs (RNC) qui sont un récent type de réseaux de neurones qui relie le traitement d’images à l’apprentissage automatique, afin de déterminer de la manière la plus précise le type tumoral.
  • Thumbnail Image
    Item
    Classification automatique des images histologiques du cancer du sein par réseaux de neurones convolutifs (RNC)
    (CERIST, 2018-08-01) Setitra, Insaf; Meziane, Insaf; Mayouf, Mouna Sabrine; Hamrioui, Amel
    Après le cancer de la peau, le cancer du sein est le deuxième type de cancer le plus commun chez la femme à l’échelle mondiale. Ce dernier enregistre un taux de mortalité assez élevé comparé aux autres types de cancer. (Spanhol, Oliveira et al. 2016). Le diagnostic des tumeurs du sein pour différencier les cellules bénignes des malignes établi par le pathologiste est le fruit d’un processus minutieux, fastidieux, long et sujet à plusieurs erreurs et divergence d’avis. Afin d’essayer de palier à ces inconvénients, un vif intérêt s’est porté sur l’automatisation du processus du diagnostic. Dans ce travail, nous reportons les différentes méthodes utilisées jusque-là par la communauté scientifique et nous exposons notre méthode basée sur la classification par réseaux de neurones convolutifs (RNC) qui sont un récent type de réseaux de neurones qui relie le traitement d’images à l’apprentissage automatique, afin de déterminer de la manière la plus précise le type tumoral.
  • Thumbnail Image
    Item
    Object classification in videos An overview
    (CERIST, 2012) Setitra, Insaf
    In this article, we will discuss the classification of moving objects in videos. An overview of classical steps in video classification will be given and a particular attention will be given to classification in video surveillance since classification in this kind of systems is very important and plays a primary role in several functions such as event classification, speed control, classification of intrusions and so forth.
  • Thumbnail Image
    Item
    Old Manuscripts Restoration Using Segmentation and Texture Inpainting
    (CERIST, 2013-09) Setitra, Insaf; Meziane, Abdelkrim
    While manifold works are done in the area of natural image processing to enhance image content, images of documents and more specifically old manuscripts images still suffer from degradations. State of the art approaches show performance in restoring degraded manuscripts images but still a lake is observed in this area. Old manuscripts images are important in that they present a part of a culture for many countries. In this paper we consider restoration of Algerian old manuscripts which present much degradation due to physical conditions and human handling. Restoration is performed by removing spurious details and inpainting missed content. We overview important works done in this area and propose a system to restore content. The system is divided in two modules; the former is concerned with elimination of ink spots, stamps and border notes. The second module is concerned with texture inpainting which repaints missed content. The system provides an interactive user interface, enabling users to choose type of restoration and its parameters. Tests are done on synthetic images and images of real manuscripts.
  • Thumbnail Image
    Item
    Perimeter Histogram based Approach for Calligraphy Classification in Ancient Manuscripts
    (IEEE Xplore, 2014-12) Setitra, Insaf; Meziane, Abdelkrim
    Manual annotation of images is usually a mandatory task in many applications where no knowledge about the image is available. In presence of huge number of images, this task becomes very tedious and prone to human errors. In this paper, we contribute in automatic annotation of ancient manuscripts by discovering manuscript calligraphy. Ancient manuscripts count a very large number of Persian and Maghrebi writing especially in Noth Africa. Distinguishing between these two calligraphies allows better classifying them and so annotating them. We use background constructing followed by extraction of simple features to classify manuscript calligraphies using an SVM classifier.
  • Thumbnail Image
    Item
    Visual Data Mining by Virtual Reality for Protein-Protein Interaction Networks
    (CERIST, 2018-03-28) Aouaa, Noureddine; Gherbi, Rachid; Meziane, Abdelkrim; Hadjar, Hayat ; Setitra, Insaf
    Currently, visualization techniques in the genetic field require a very important modeling phase in terms of resources. Traditional modeling techniques (in two dimensions) are rarely adapted to manage and process this mass of information. To overcome this kind of problem, we propose to use a new graph modeling technique that, used in conjunction with the concept of virtual reality, allows biologists to have a wide visibility through several points of view, thus facilitating them the exploration of massive data. The general principle of our approach is to build a biological network model in the form of a graph with a spatial representation adapted to the visualization of biological networks in a virtual environment. The results show that the improvement of the node distribution algorithm allows a better and more intuitive visualization, compared to the equivalent two-dimensional representations.
  • Thumbnail Image
    Item
    WebVR based Interactive Visualization of Open Health Data
    (CERIST, 2018-04-22) Hadjar, Hayet; Meziane, Abdelkrim; Guerbi, Rachid; Setitra, Insaf; Zeghichi, Seyf eddine; Lahmil, Abdessalam
    Visualization and manipulation of complex and multivariate data in virtual worlds is important for both holders of these data and for their users. Indeed, Virtual Reality helps to make multidimensional data more intelligible and to bring useful information and knowledge. Offering Virtual Reality to browsers, also known as Web Virtual Reality, moreover, simplifies access, multi sharing and manipulation of complex and multivariate data. In this paper, we propose a new system for exploring and visualizing multidimensional data in WebVR. The system implements three methods that we compare based upon several criteria. As the area of application; we use health data that we collect from open data portals.