Centre dedocumentation
scientifique

Initialization methods of convolutional neural networks for detection of image manipulations / Ivan Castillo Camacho (2021)  

Public Titre : Initialization methods of convolutional neural networks for detection of image manipulations Titre original : Méthodes d'initialisation des réseaux de neurones convolutifs pour la détection des manipulations d'images Type de document : Thèse/HDR Auteurs : Ivan Castillo Camacho, Auteur ; Kai Wang, Directeur de thèse Editeur : Grenoble [France] : Université Grenoble Alpes Année de publication : 2021 Importance : 145 p. Format : 21 x 30 cm Note générale : bibliographie Thèse pour obtenir le grade de Docteur de l'Université Grenoble, spécialité : signal, image, paroles, télécoms Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image optique [Termes IGN] altération [Termes IGN] apprentissage profond [Termes IGN] classification par réseau neuronal convolutif [Termes IGN] covariance [Termes IGN] détection d'anomalie [Termes IGN] estompage [Termes IGN] filtre passe-haut [Termes IGN] flux de données [Termes IGN] infraction [Termes IGN] manipulation de données [Termes IGN] qualité des données [Termes IGN] retouche [Termes IGN] variance Index. décimale : THESE Thèses et HDR Résumé : (auteur) Fake images and videos have engulfed mass communication media. This is not something recent, manipulations and forgeries have occurred since the advent of photography itself. These alterations can go from innocent retouches in an attempt to make an image visually attractive to the spread of misleading information or even the use of false media in legal instances. Accordingly, the creation of methods that can help us assure the authenticity of an image presented as non-modified is of paramount importance. In this thesis, we aim at detecting image manipulation operations using deep learning techniques. We present three methods showing the progression of our work under one common objective, i.e, the design and test of Convolutional Neural Network (CNN) initialization methods for image forensic problems with a variance stability focus for the output of a CNN layer.First, we carry out an extensive review of the state of the art in deep-learning-based methods for image forensics. From this review we can confirm that the first layer of a CNN has big impact on the final performance. Specifically, the initialization used on the first-layer filters plays an important role that should be in line with the image forensic task in hand.As our first attempt to address this research problem, we propose a low-complexity initialization method for CNNs. Taking advantage of previous methods designed for the computer vision field, we extend the popular Xavier method to design a filter that would provide variance stability after a convolution operation. This method generates a set of random high-pass filters for the initialization of a CNN's first layer. These filters allow us to better identify forensic traces which usually lie towards the high-frequency part of the image.This first approach constitutes a good staring point of our work. However, a wrong assumption, largely utilized in the research community, was made. This is corrected in our second method where we follow a different data-dependent approach and take into consideration the real statistical properties of natural images. Accordingly, we propose a scaling method for first-layer filters which can cope well with different CNN initialization algorithms. The objective remains in keeping the stability of the variance of data flow in a CNN. We also present theoretical and experimental studies on the output variance for convolutional filter, which are the basis of our proposed data-dependent scaling.Next we describe a revisited version of our first proposal now with a corrected assumption on the statistics of natural images. More precisely, we propose an improved random high-pass initialization method which does not explicitly compute the statistics of input data. We believe that such a ``data-independent'' approach has higher flexibility and broader application range than our second method in situations where the computation of input statistics is not possible.Our proposed methods are tested over several image forensic problems and different CNN architectures.Finally, during all this thesis work we took part in a challenge competition of image forgery detection organized by the French National Research Agency and the French Directorate General of Armaments. We explain in the Appendix the objectives of the challenge along with a brief description of our work conducted for the competition. Note de contenu : 1- Introduction 2- Background knowledge and state of the art 3- Random high-pass initialization 4- Data-dependent initialization 5- Revisiting the random high-pass initialization 6- Conclusions and perspectives Numéro de notice : 28437 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Thèse française Note de thèse : Thèse de Doctorat : signal, image, paroles, télécoms : Grenoble : 2021 DOI : sans En ligne : https://tel.archives-ouvertes.fr/tel-03346063/document Format de la ressource électronique : URL Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=98833

Unsupervised object-based differencing for land-cover change detection / Jinxia Zhu in Photogrammetric Engineering & Remote Sensing, PERS, vol 83 n° 3 (March 2017)  

Public [article]  inPhotogrammetric Engineering & Remote Sensing, PERS > vol 83 n° 3 (March 2017) . - pp 225 - 236 Titre : Unsupervised object-based differencing for land-cover change detection Type de document : Article/Communication Auteurs : Jinxia Zhu, Auteur ; Yanjun Su, Auteur ; Qinghua Guo, Auteur ; Thomas C. Harmon, Auteur Année de publication : 2017 Article en page(s) : pp 225 - 236 Note générale : Bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image optique [Termes IGN] algorithme espérance-maximisation [Termes IGN] altération [Termes IGN] autocorrélation [Termes IGN] changement d'occupation du sol [Termes IGN] Chine [Termes IGN] classification non dirigée [Termes IGN] classification orientée objet [Termes IGN] détection de changement [Termes IGN] image multitemporelle [Termes IGN] image SPOT-HRV [Termes IGN] occupation du sol [Termes IGN] traitement d'image Résumé : (Auteur) One main problem of the spectral decomposition-based change detection method is the lack of efficient automatic techniques for developing the difference image. Traditional techniques generally assume that gray-level values in a difference image are independent and multitemporal images are co-registered/rectified perfectly without error. However, such assumptions are often violated because of the inevitable image misregistration and the interference of correlations between spectral bands. This study proposes an automated method based on the object-based multivariate alteration detection/maximum autocorrelation factor approach and the Gaussian mixture model-expectation maximization algorithm to obtain unsupervised difference images. This procedure is applied to bi-temporal (2005 and 2006) SPOT-HRV images at Panyu District Ponds, China. Results show that the proposed method successfully excludes the correlations of spectral bands and the influence of misregistration, as evidenced by a higher accuracy (up to 93.6 percent). These unique technical characteristics make this analytical framework suitable for detecting changes. Numéro de notice : A2017-089 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Article DOI : 10.14358/PERS.83.3.225 En ligne : https://doi.org/10.14358/PERS.83.3.225 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=84424 [article]

Réservation

Réserver ce document

Exemplaires (2)

Code-barres	Cote	Support	Localisation	Section	Disponibilité
105-2017031	SL	Revue	Centre de documentation	Revues en salle	Disponible
105-2017032	SL	Revue	Centre de documentation	Revues en salle	Disponible

A new computerized ionosphere tomography model using the mapping function and an application to the study of seismic-ionosphere disturbance / Jian Kong in Journal of geodesy, vol 90 n° 8 (August 2016)  

Public [article]  inJournal of geodesy > vol 90 n° 8 (August 2016) . - pp 741 – 755 Titre : A new computerized ionosphere tomography model using the mapping function and an application to the study of seismic-ionosphere disturbance Type de document : Article/Communication Auteurs : Jian Kong, Auteur ; Yi Bin Yao, Auteur ; Lei Liu, Auteur ; et al., Auteur Année de publication : 2016 Article en page(s) : pp 741 – 755 Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Applications de géodésie spatiale [Termes IGN] altération [Termes IGN] correction troposphérique [Termes IGN] International Reference Ionosphere [Termes IGN] Japon [Termes IGN] séisme [Termes IGN] teneur totale en électrons [Termes IGN] tomographie [Termes IGN] troposphère Résumé : (auteur) A new algorithm for ionosphere tomography using the mapping function is proposed in this paper. First, the new solution splits the integration process into four layers along the observation ray, and then, the single-layer model (SLM) is applied to each integration part using a mapping function. Next, the model parameters are estimated layer by layer with the Kalman filtering method by introducing the scale factor (SF) γ to solve the ill-posed problem. Finally, the inversed images of different layers are combined into the final CIT image. We utilized simulated data from 23 IGS GPS stations around Europe to verify the estimation accuracy of the new algorithm; the results show that the new CIT model has better accuracy than the SLM in dense data areas and the CIT residuals are more closely grouped. The stability of the new algorithm is discussed by analyzing model accuracy under different error levels (the max errors are 5TECU, 10TECU, 15TECU, respectively). In addition, the key preset parameter, SFγ, which is given by the International Reference Ionosphere model (IRI2012). The experiment is designed to test the sensitivity of the new algorithm to SF variations. The results show that the IRI2012 is capable of providing initial SF values. Also in this paper, the seismic-ionosphere disturbance (SID) of the 2011 Japan earthquake is studied using the new CIT algorithm. Combined with the TEC time sequence of Sat.15, we find that the SID occurrence time and reaction area are highly related to the main shock time and epicenter. According to CIT images, there is a clear vertical electron density upward movement (from the 150-km layer to the 450-km layer) during this SID event; however, the peak value areas in the different layers were different, which means that the horizontal movement velocity is not consistent among the layers. The potential physical triggering mechanism is also discussed in this paper. Compared with the SLM, the RMS of the new CIT model is improved by 16.78%, while the CIT model could provide the three-dimensional variation in the ionosphere. Numéro de notice : A2016-504 Affiliation des auteurs : non IGN Thématique : POSITIONNEMENT Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.1007/s00190-016-0906-9 En ligne : http://dx.doi.org/10.1007/s00190-016-0906-9 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=81528 [article]

Practical geodesy: part 5 chart projections / Huibert-Jan Lekkerkerk in Geoinformatics, vol 10 n° 6 (01/09/2007)

Public [article]  inGeoinformatics > vol 10 n° 6 (01/09/2007) . - pp 48 - 51 Titre : Practical geodesy: part 5 chart projections Type de document : Article/Communication Auteurs : Huibert-Jan Lekkerkerk, Auteur Année de publication : 2007 Article en page(s) : pp 48 - 51 Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Projections [Termes IGN] altération [Termes IGN] ellipsoïde (géodésie) [Termes IGN] projection [Termes IGN] projection Universal Transverse Mercator [Termes IGN] système de coordonnées Résumé : (Auteur) In the previous article, the importance of chart projections was discribed. An important aspect of projecting information from an ellipsoid on a flat surface is that the information will be distorded. In this article, some of these distorsions and their effect will be described. Copyright GEOinformatics Numéro de notice : A2007-419 Affiliation des auteurs : non IGN Thématique : GEOMATIQUE/POSITIONNEMENT Nature : Article DOI : sans Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=28782 [article]

Voir aussi

• Practical geodesy: part 4 chart projections (1) / Huibert-Jan Lekkerkerk in Geoinformatics, vol 10 n° 5 (01/07/2007)
• Practical geodesy: part 6 software set-up / Huibert-Jan Lekkerkerk in Geoinformatics, vol 10 n° 7 (01/11/2007)

Réservation

Réserver ce document

Exemplaires (1)

Code-barres	Cote	Support	Localisation	Section	Disponibilité
262-07061	SL	Revue	Centre de documentation	Revues en salle	Disponible

The technique of distance decayed visibility for forest landscape visualization / C. Kumsap in International journal of geographical information science IJGIS, vol 19 n° 6 (july 2005)  

Public [article]  inInternational journal of geographical information science IJGIS > vol 19 n° 6 (july 2005) . - pp 723 - 744 Titre : The technique of distance decayed visibility for forest landscape visualization Type de document : Article/Communication Auteurs : C. Kumsap, Auteur ; F. Borne, Auteur ; D. Moss, Auteur Année de publication : 2005 Article en page(s) : pp 723 - 744 Note générale : Bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Systèmes d'information géographique [Termes IGN] altération [Termes IGN] distance de propagation [Termes IGN] forêt [Termes IGN] paysage [Termes IGN] point d'appui [Termes IGN] système d'information géographique [Termes IGN] visualisation 3D Résumé : (Auteur) The paper discusses an approach known as 'distance-decayed visibility', implemented using GIS, to the composition of 3D scenes for static forest landscape visualization. The distance-decay concept is employed to model a spatial interaction in terms of visibility between any observation points and vegetations visible from those points. The vegetation visibility is measured in level of detail where high visibility uses high level of detail and vice versa. The interaction is modelled mathematically using an exponential decay function. The point distribution of projected models on a 3D scene is calculated in GIS to assign an appropriate value of 'visibility decay constant' to relate the distance from the observation point to the vegetations with the level of detail of the vegetations in the function. The advantages of the technique as measured against the quantitative benchmarks of rendering time and the number of polygons computed are addressed. Although the forest landscape of the study area consists of a large number of vegetation models, the technique helps to model the relation of vegetation visibility and viewing distance so realistically that the technique yields the visual quality of the result 3D scene visually and statistically comparable to the finest resolution scene. Numéro de notice : A2005-290 Affiliation des auteurs : non IGN Thématique : FORET/GEOMATIQUE/INFORMATIQUE Nature : Article DOI : 10.1080/13658810500104880 En ligne : https://doi.org/10.1080/13658810500104880 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=27426 [article]

Réservation

Réserver ce document

Exemplaires (2)

Code-barres	Cote	Support	Localisation	Section	Disponibilité
079-05061	RAB	Revue	Centre de documentation	En réserve 3L	Disponible
079-05062	RAB	Revue	Centre de documentation	En réserve 3L	Disponible

Fractal dimension and inconsistencies in cartographic line representations / Jean-Claude Müller in Cartographic journal (the), vol 23 n° 2 (December 1986)  

Permalink

Orthophotographie, modèles numériques de terrain / Patrice Denis (05/06/1979)

Permalink

La projection conique conforme de Lambert / Albert Reyt (1950)

Permalink

Mathématiques appliquées : éléments de la théorie des surfaces, 2. Figures / J. Laborde (1949)

Permalink

Notions de géodésie théorique et pratique, 1. 1ère partie / M. Lejeune (1948)

Permalink

Notions de géodésie théorique et pratique, 2. 2ème partie / M. Lejeune (1948)

Permalink

Notions de géodésie théorique et pratique, 3. 3ème partie / M. Lejeune (1948)

Permalink

Mathématiques appliquées : éléments de la théorie des surfaces, 1. Texte / J. Laborde (1946)

Permalink

Traité des projections des cartes géographiques à l'usage des cartographes et géodésiens, 4. Fasc 4 Théorie de la représentation conforme, emploi des projections rigoureusement conformes en géodésie / Ludovic Driencourt (1932)

Permalink