20 sep

diff --git a/DOCS/CSD-01-1133.pdf b/DOCS/CSD-01-1133.pdf
new file mode 100644 (file)
index 0000000..c988821
Binary files /dev/null and b/DOCS/CSD-01-1133.pdf differ

diff --git a/DOCS/bresenham_line.pdf b/DOCS/bresenham_line.pdf
new file mode 100644 (file)
index 0000000..f5592f5
Binary files /dev/null and b/DOCS/bresenham_line.pdf differ

diff --git a/DOCS/ct-reconstruction-cuda.pdf b/DOCS/ct-reconstruction-cuda.pdf
new file mode 100644 (file)
index 0000000..f1013f0
Binary files /dev/null and b/DOCS/ct-reconstruction-cuda.pdf differ

diff --git a/DOCS/iccv2009.pdf b/DOCS/iccv2009.pdf
new file mode 100644 (file)
index 0000000..35cbec5
Binary files /dev/null and b/DOCS/iccv2009.pdf differ

diff --git a/DOCS/paper_cuda.pdf b/DOCS/paper_cuda.pdf
new file mode 100644 (file)
index 0000000..6801f67
Binary files /dev/null and b/DOCS/paper_cuda.pdf differ

diff --git a/DOCS/snake-gvf-opencl-2012.pdf b/DOCS/snake-gvf-opencl-2012.pdf
new file mode 100644 (file)
index 0000000..664acde
Binary files /dev/null and b/DOCS/snake-gvf-opencl-2012.pdf differ

diff --git a/DOCS/snake-gvf-poumons.pdf b/DOCS/snake-gvf-poumons.pdf
new file mode 100644 (file)
index 0000000..d33c4a9
Binary files /dev/null and b/DOCS/snake-gvf-poumons.pdf differ

diff --git a/DOCS/snake_gvf_gpu_2012.pdf b/DOCS/snake_gvf_gpu_2012.pdf
new file mode 100644 (file)
index 0000000..ecd139e
Binary files /dev/null and b/DOCS/snake_gvf_gpu_2012.pdf differ

diff --git a/DOCS/snake_gvf_gpu_opencl.bib b/DOCS/snake_gvf_gpu_opencl.bib
new file mode 100644 (file)
index 0000000..d6a2773
--- /dev/null
+++ b/DOCS/snake_gvf_gpu_opencl.bib
@@ -0,0 +1,13 @@
+@article{
+year={2012},
+issn={1861-8200},
+journal={Journal of Real-Time Image Processing},
+doi={10.1007/s11554-012-0257-6},
+title={Real-time gradient vector flow on GPUs using OpenCL},
+url={http://dx.doi.org/10.1007/s11554-012-0257-6},
+publisher={Springer-Verlag},
+keywords={Gradient Vector Flow; GPU; OpenCL},
+author={Smistad, Erik and Elster, AnneC. and Lindseth, Frank},
+pages={1-8},
+language={English}
+}

diff --git a/THESE/Chapters/chapter2/chapter2.tex b/THESE/Chapters/chapter2/chapter2.tex
index b02fcc477106b63d58ef80b35a1f83645d39e7c0..9d12c732fc487b1dea6ce71955859b6f5f76be77 100644 (file)
--- a/THESE/Chapters/chapter2/chapter2.tex
+++ b/THESE/Chapters/chapter2/chapter2.tex
@@ -508,19 +508,46 @@ Récemment, Xiao et Liu ont décrit dans \cite{xiao2010efficient} une implément
 Dès 2003, on recense d'importants travaux liés à l'imagerie médicale mettant en \oe uvre des algorithmes \textit{level set} sur GPU. C'est le cas de \cite{lefohn2003inter,lefohn2003interactive} où les auteurs décrivent une solution de visualisation des coupes d'une mesure volumique réalisés par résonnance magnétique (IRM) en exploitant pour la première fois le caractère creux du système d'équations à résoudre, \textit{i.e.} variante narrow-band, contrairement à la première solution 2D présentée dans \cite{rumpf2001level} qui implémente la version standard. En ne transférant au GPU, pour chaque itération, que les petits pavés de données actifs et en les  rangeant alors de manière contigue en texture pour optimiser les accès en lecture, les auteurs sont ainsi parvenu à effectuer, pour des données volumiques de 256$\times$256$\times$175, entre 3.5 et 70 itérations par seconde, à comparer aux 50 itérations par seconde en 2D sur image de 128$^2$ pixels otenues dans \cite{rumpf2001level}. La limitation principale de cettesolution est celle des dimensions maximales admises pour une texture qui était de 2048$^2$ pour le GPU ATI Radeon 9800 pro employé (et programmé en openGL, car ni openCL ni CUDA n'étaient encore disponible à l'époque).
 Les autres solutions GPU proposées depuis sont également basées sur la variante \textit{narrow-band} (bande étroite) des \textit{level-set} \cite{lefohn2005streaming,cates2004gist,jeong2009scalable}, mais seule \cite{jeong2009scalable} s'affranchit des transferts CPU/GPU à chaque itération pour déterminer et transférer les pavés actifs. La solution retenue est d'employer les opérations atomiques pour assurer l'accès exclusif à la liste des pavés en mémoire GPU. Cela permet de descendre à 3~ms par itération pour une image de 512$^2$ pixels.
 
-La plus performante des implémentations à ce jour est celle décrite dans \cite{Roberts:2010:WGA:1921479.1921499} qui parvient à des itérations dont la durée varie, sur GTX280,  de 1.8 à 6.5~ms pour des données volumiques de 256$^3$ pixels issues d'examen IRM, pour une moyenne de 3.2~ms sur les 2200 itérations de l'exemple fourni (cerveau en 7~s, Figure \ref{fig-l7-brain}). Une optimisation poussée y a été effectuée pour rendre l'algorithme efficace, en particulier au travers de la refonte du code responsable de la détermination des pavés actifs. Il parvient cette fois à déterminer l'ensemble minimal de pavés actifs et à rendre cette détermination efficace sur le GPU en gérant parallèlement plusieurs tampons, chacun associé à une direction particulière en 6-connexité. Une étape de résolution des doublons est ensuite effectuée avant de les compacter de manière contigue comme cele était déjà fait dans \cite{lefohn2003inter}. Toutefois, tenir à jour cette liste de pavés représente encore 77\% du temps de calcul après cette optimisation.
-%TODO dire qu'il n'utilise pas de shmem !
+La plus performante des implémentations à ce jour est celle décrite dans \cite{Roberts:2010:WGA:1921479.1921499} qui parvient à des itérations dont la durée varie, sur GTX280,  de 1.8 à 6.5~ms pour des données volumiques de 256$^3$ pixels issues d'examen IRM, pour une moyenne de 3.2~ms sur les 2200 itérations de l'exemple fourni (cerveau en 7~s, Figure \ref{fig-l7-brain}). Une optimisation poussée y a été effectuée pour rendre l'algorithme efficace, en particulier au travers de la refonte du code responsable de la détermination des pavés actifs. Il parvient cette fois à déterminer l'ensemble minimal de pavés actifs et à rendre cette détermination efficace sur le GPU en gérant parallèlement plusieurs tampons, chacun associé à une direction particulière en 6-connexité. Une étape de résolution des doublons est ensuite effectuée avant de les compacter de manière contigue comme cela était déjà fait dans \cite{lefohn2003inter}.Tout cela est réalisé sans recourir à la mémoire partagée qui s'avère complexe voire impossile à utiliser efficacement lorsque les éléments à accéder sont très irrégulièrement répartis en mémoire. 
+
 Ce faisant, le nombre cumulé total de pavés ainsi traités lors des 2200 itérations de la segmentation der l'image d'exemple s'élève à 294 millions à comparer aux 4877 millions traités par l'algorithme \textit{narrow-band} standard. Il est à noter que la durée d'exécution d'une itération dans cette variante dépend plus fortement de la proportion de pavés actifs que pour \textit{narrow-band} standard. Les deux courbes sont globalement affines et se croisent pour une proportion de pavés actifs proche de 10\%.
-Cela peut représenter une piste pour une optimisation supplémentaire qui ne semble pas su justifier avec l'image et l'initialisation dont les performances sont détaillées, mais qui pourrait l'être dans d'autres conditions, comme peut le suggérer le temps de segmentation de 16~s nécessaire pour l'image des reins (Figure \ref{fig-l7-reins}) et de l'aorte, malgré des dimensions comparables.
+Si l'on considère que malgré les stratégies adoptées, tenir à jour cette liste de pavés représente encore 77\% du temps de calcul, cela peut représenter une piste pour une optimisation supplémentaire qui ne semble pas su justifier avec l'image et l'initialisation dont les performances sont détaillées, mais qui pourrait l'être dans d'autres conditions, comme peut le suggérer le temps de segmentation de 16~s nécessaire pour l'image des reins (Figure \ref{fig-l7-reins}) et de l'aorte, aux dimensions comparables.
 
 \begin{figure}
   \centering
 \subfigure[Cerveau 256$\times$256$\times$256 en 7~s]{\label{fig-l7-brain}\includegraphics[height=4cm]{/home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/l7-brain7s.png}}\quad
 \subfigure[Reins et aorte, 256$\times$256$\times$272 en 16~s]{\label{fig-l7-reins}\includegraphics[height=4cm]{/home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/l7-reins16s.png}}
 \caption{Segmentation d'images issues d'examens IRM par la méthode des level set à bande étroite.}
-\label{fig-meanshift-castle}
+\label{fig-l7-narrow}
 \end{figure}
 
+Les algorithmes de type \textit{snake}, très coûteux en temps de calcul, pouvaient prétendre à bénéficier largement de la technologie des GPU pour améliorer leurs performances, mais seule la variante paramétrique GVF à véritablement été implémentée de manière spécifique et efficace \cite{snakegvf06, bauer2009segmentation, li2011robust, snakegvfopencl12}. Les variantes de type géométrique, principalement en raison de l'irrégularité des motifs d'accès à la mémoire, restent à ce jour sans implémentation GPU.
+Parmi les premières solutions décrites, \cite{snakegvf06} propose une implémentation réalisée en openGL, où les données de gradient sont compactées en texture RVBA de manière à s'affranchir du format 16 bits de la représentation : les deux premiers canaux R et V contiennent les valeursreprésentant respectivement le gradients selon $dx$ et $dy$ sous une forme codée par la valeurs des 2 autres canaux. 
+Par ailleurs, une approximation du système linéaire à résoudre est proposée afin de donner une structure bande symétrique à la matrice à inverser, ce qui améliore considérablement l'efficacité des accès aux données au travers du cache.
+
+Les performances annoncées montrent tout d'abord que l'approximation adoptée n'a qu'un impact extrêmement limité sur le résulat de la segmentation avec un écart radial maximal inférieur à 1.3 pixel par rapport au calcul exact effectué sur CPU. Enfin, la segmentation de l'image d'exemple en 1024$^2$ pixels s'effectue en un total de 11~s après l'initialisation manuelle reproduite à la figure \ref{fig-snakegvf}. Cela est annoncé comme presque 30 fois plus rapide que l'implémentation CPU de référence, mais demeure beaucoup trop lent pour un usage interactif.
+
+Une solution directe employant la transformée de fourier pour inverser le système à résoudre a été décrite récemment dans  \cite{zheng2012fast}et programmée en employant la bibliothèque openGL. Les exemples fournis montrent des objets segmentés dans des images d'environ 10000 pixels en une durée de l'ordre de la demi seconde.   
+
+En adaptant sur GPU une variante dite FD-snake \cite{li2011robust} du snake GVF (pour Fourier Descriptors) permettant une convergence plus rapide et un calcul parallèle beaucoup plus adapté au GPU, Li \textit{et al.} parviennent quant à eux à suivre les déformations d'un contour en temps réel dans des images issues d'examens échographique ; Un contour de 100 points pouvant converger convenablement en à peine 30~ms. Une contribution supplémentaire de cette implémentation est de permettre une initialisation simplifiée et semi-automatique du contour. 
+
+La plus aboutie des implémentations actuelles du snake GVF est enfin celle présentée par Smistad \textit{et al.} dans \cite{snakegvfopencl12} et où les auteurs ont concentré leur effort sur l'optimisation des accès mémoire lors du calcul du GVF. Ils ont comparé 8 combinaisons possibles impliquant l'emploi des mémoires partagée et de texture ainsi que la représentation des nombres selon le format classique 32 bits ou selon un format compressé sur 16 bits. Il en ressort que l'association la plus performante est celle des textures et du format de données sur 16 bits.
+Les performances sont alors nettement en hausse avec des segmentations d'images médicales d'IRM de 512$^2$ pixels effectuées en 41~ms sur Nvidia C2070 et 28~ms sur ATI 5870 (512 itérations). L'implémentation réalisée en openGL permet d'exécuter le code sur les GPU des deux principaux fabricants.   
+
+\subsection{Algorithmes hybrides}
+Le détecteur de contour \textit{gPb} décrit dans \cite{arbelaez2011contour} et que l'on considère comme la référence actuelle pour la semgentation d'objets et personnages dans des image naturelles, à été implémenté en CUDA par Catanzaro \textit{et al.} et est décrit dans \cite{5459410}. La qualité des contours extraits y est préservée et le temps de traitement y est réduit d'un facteur supérieur à 100 : les contours des images de 0.15~MP de la base de test BSDS \cite{martin2001database} sont ainsi traitées en 2 secondes environ sur GPU C1060.
+L'apport principal de ces travaux réside dans la solution conçue pour le calcul des histogrammes locaux, qui dans l'algorithme original s'étendaient sur des demi-disques centrés sur chaque pixel. La parallélisation réalisée fait l'approximation de chaque demi-disque en un rectangle de même surface dont un des grands cotés à le centre du disque pour milieu. Les rectangles sont ensuite pivotés par une rotation basée sur la discrétisation de Bresenham \cite{bresenham1965algorithm} pour en aligner les cotés avec les cotés de l'image et pouvoir employer la technique des images cumulées pour calculer rapidement l'histogramme.   
+La figure \ref{fig-gPb} présente quelques résultats d'extraction de contours.
+\begin{figure}
+  \centering
+\includegraphics[height=4cm]{/home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/gPb_examples.png}
+\caption{Extraction de contour par la version GPU de l'algorithme gPb. Les images sont issues de la base BSDS  \cite{martin2001database}}
+\label{fig-gPb}
+\end{figure}
+
+     
+
+
 
 
 

diff --git a/THESE/Chapters/chapter2/img/gPb_examples.png b/THESE/Chapters/chapter2/img/gPb_examples.png
new file mode 100644 (file)
index 0000000..f6084e1
Binary files /dev/null and b/THESE/Chapters/chapter2/img/gPb_examples.png differ

diff --git a/THESE/biblio.bib b/THESE/biblio.bib
index 850c5b55c13b81250a6127276d1b2abf470ad644..92f642f2b1cc1113d3fb01a94a0ec6d12242d009 100755 (executable)
--- a/THESE/biblio.bib
+++ b/THESE/biblio.bib
@@ -1226,7 +1226,7 @@ pages={166-175}
   year={2011},
   organization={IEEE}
 }
-@incollection{
+@incollection{snakegvf06,
 year={2006},
 isbn={978-3-540-48628-2},
 booktitle={Advances in Visual Computing},
@@ -1240,3 +1240,59 @@ publisher={Springer Berlin Heidelberg},
 author={He, Zhiyu and Kuester, Falko},
 pages={191-201}
 }
+@article{snakegvfopencl12,
+year={2012},
+issn={1861-8200},
+journal={Journal of Real-Time Image Processing},
+doi={10.1007/s11554-012-0257-6},
+title={Real-time gradient vector flow on GPUs using OpenCL},
+url={http://dx.doi.org/10.1007/s11554-012-0257-6},
+publisher={Springer-Verlag},
+keywords={Gradient Vector Flow; GPU; OpenCL},
+author={Smistad, Erik and Elster, AnneC. and Lindseth, Frank},
+pages={1-8},
+language={English}
+}
+@inproceedings{bauer2009segmentation,
+  title={Segmentation of airways based on gradient vector flow},
+  author={Bauer, Christian and Bischof, Horst and Beichel, Reinhard},
+  booktitle={International workshop on pulmonary image analysis, Medical image computing and computer assisted intervention},
+  pages={191--201},
+  year={2009},
+}
+@article{zheng2012fast,
+  title={A Fast GVF Snake Algorithm on the GPU},
+  author={Zheng, Zuoyong and Zhang, Ruixia},
+  journal={image},
+  volume={2},
+  pages={4},
+  year={2012}
+}
+@INPROCEEDINGS{5459410, 
+author={Catanzaro, Bryan and Bor-Yiing Su and Sundaram, N. and Yunsup Lee and Murphy, Mark and Keutzer, K.}, 
+booktitle={Computer Vision, 2009 IEEE 12th International Conference on}, 
+title={Efficient, high-quality image contour detection}, 
+year={2009}, 
+pages={2381-2388}, 
+keywords={Application software;Computer vision;Detection algorithms;Detectors;Image analysis;Image segmentation;Object detection;Object recognition;Parallel algorithms;Runtime}, 
+doi={10.1109/ICCV.2009.5459410}, 
+ISSN={1550-5499},}
+@article{bresenham1965algorithm,
+  title={Algorithm for computer control of a digital plotter},
+  author={Bresenham, Jack E},
+  journal={IBM Systems journal},
+  volume={4},
+  number={1},
+  pages={25--30},
+  year={1965},
+  publisher={IBM}
+}
+@inproceedings{martin2001database,
+  title={A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics},
+  author={Martin, David and Fowlkes, Charless and Tal, Doron and Malik, Jitendra},
+  booktitle={Computer Vision, 2001. ICCV 2001. Proceedings. Eighth IEEE International Conference on},
+  volume={2},
+  pages={416--423},
+  year={2001},
+  organization={IEEE}
+}
\ No newline at end of file

diff --git a/THESE/these.aux b/THESE/these.aux
index 84c3d9afba4ba36fb9facd5c09338c537fccf07d..fc1f3afce37ffcf06b8501393d67c68026b2c742 100644 (file)
--- a/THESE/these.aux
+++ b/THESE/these.aux
@@ -262,28 +262,45 @@
 \@writefile{lof}{\contentsline {subfigure}{\numberline{(a)}{\ignorespaces {Image originale}}}{35}{figure.2.16}}
 \@writefile{lof}{\contentsline {subfigure}{\numberline{(b)}{\ignorespaces {Image segment\IeC {\'e}e par mean-shift standard}}}{35}{figure.2.16}}
 \@writefile{lof}{\contentsline {subfigure}{\numberline{(c)}{\ignorespaces {Image segment\IeC {\'e}e par mean-shift kd-tree}}}{35}{figure.2.16}}
+\citation{snakegvf06}
+\citation{bauer2009segmentation}
+\citation{li2011robust}
+\citation{snakegvfopencl12}
+\citation{snakegvf06}
+\citation{zheng2012fast}
 \newlabel{fig-l7-brain}{{2.17(a)}{36}{Subfigure 2 2.17(a)\relax }{subfigure.2.17.1}{}}
 \newlabel{sub@fig-l7-brain}{{(a)}{36}{Subfigure 2 2.17(a)\relax }{subfigure.2.17.1}{}}
 \newlabel{fig-l7-reins}{{2.17(b)}{36}{Subfigure 2 2.17(b)\relax }{subfigure.2.17.2}{}}
 \newlabel{sub@fig-l7-reins}{{(b)}{36}{Subfigure 2 2.17(b)\relax }{subfigure.2.17.2}{}}
 \@writefile{lof}{\contentsline {figure}{\numberline {2.17}{\ignorespaces Segmentation d'images issues d'examens IRM par la m\IeC {\'e}thode des level set \IeC {\`a} bande \IeC {\'e}troite.}}{36}{figure.2.17}}
-\newlabel{fig-meanshift-castle}{{2.17}{36}{Segmentation d'images issues d'examens IRM par la méthode des level set à bande étroite}{figure.2.17}{}}
+\newlabel{fig-l7-narrow}{{2.17}{36}{Segmentation d'images issues d'examens IRM par la méthode des level set à bande étroite}{figure.2.17}{}}
 \@writefile{lof}{\contentsline {subfigure}{\numberline{(a)}{\ignorespaces {Cerveau 256$\times $256$\times $256 en 7~s}}}{36}{figure.2.17}}
 \@writefile{lof}{\contentsline {subfigure}{\numberline{(b)}{\ignorespaces {Reins et aorte, 256$\times $256$\times $272 en 16~s}}}{36}{figure.2.17}}
-\@writefile{toc}{\contentsline {chapter}{\numberline {3}La segmentation orient\IeC {\'e}e r\IeC {\'e}gions dans les images bruit\IeC {\'e}es}{37}{chapter.3}}
+\citation{li2011robust}
+\citation{snakegvfopencl12}
+\citation{arbelaez2011contour}
+\citation{5459410}
+\citation{martin2001database}
+\citation{bresenham1965algorithm}
+\citation{martin2001database}
+\citation{martin2001database}
+\@writefile{toc}{\contentsline {subsection}{\numberline {2.6.5}Algorithmes hybrides}{37}{subsection.2.6.5}}
+\@writefile{lof}{\contentsline {figure}{\numberline {2.18}{\ignorespaces Extraction de contour par la version GPU de l'algorithme gPb. Les images sont issues de la base BSDS \cite  {martin2001database}}}{38}{figure.2.18}}
+\newlabel{fig-gPb}{{2.18}{38}{Extraction de contour par la version GPU de l'algorithme gPb. Les images sont issues de la base BSDS \cite {martin2001database}\relax }{figure.2.18}{}}
+\@writefile{toc}{\contentsline {chapter}{\numberline {3}La segmentation orient\IeC {\'e}e r\IeC {\'e}gions dans les images bruit\IeC {\'e}es}{39}{chapter.3}}
 \@writefile{lof}{\addvspace {10\p@ }}
 \@writefile{lot}{\addvspace {10\p@ }}
-\@writefile{toc}{\contentsline {section}{\numberline {3.1}Pr\IeC {\'e}sentation - existant}{37}{section.3.1}}
-\@writefile{toc}{\contentsline {section}{\numberline {3.2}La parall\IeC {\`e}lisation du snake polygonal}{37}{section.3.2}}
-\@writefile{toc}{\contentsline {chapter}{\numberline {4}Le filtrage des images sur GPU}{39}{chapter.4}}
+\@writefile{toc}{\contentsline {section}{\numberline {3.1}Pr\IeC {\'e}sentation - existant}{39}{section.3.1}}
+\@writefile{toc}{\contentsline {section}{\numberline {3.2}La parall\IeC {\`e}lisation du snake polygonal}{39}{section.3.2}}
+\@writefile{toc}{\contentsline {chapter}{\numberline {4}Le filtrage des images sur GPU}{41}{chapter.4}}
 \@writefile{lof}{\addvspace {10\p@ }}
 \@writefile{lot}{\addvspace {10\p@ }}
-\@writefile{toc}{\contentsline {section}{\numberline {4.1}Algorithme de r\IeC {\'e}duction de bruit par recherche des lignes de niveaux}{39}{section.4.1}}
-\@writefile{toc}{\contentsline {section}{\numberline {4.2}Filtre m\IeC {\'e}dian}{39}{section.4.2}}
-\@writefile{toc}{\contentsline {section}{\numberline {4.3}Filtres de convolution}{39}{section.4.3}}
+\@writefile{toc}{\contentsline {section}{\numberline {4.1}Algorithme de r\IeC {\'e}duction de bruit par recherche des lignes de niveaux}{41}{section.4.1}}
+\@writefile{toc}{\contentsline {section}{\numberline {4.2}Filtre m\IeC {\'e}dian}{41}{section.4.2}}
+\@writefile{toc}{\contentsline {section}{\numberline {4.3}Filtres de convolution}{41}{section.4.3}}
 \bibstyle{plain}
 \bibdata{biblio}
-\@writefile{toc}{\contentsline {chapter}{\numberline {5}Conclusion g\IeC {\'e}n\IeC {\'e}rale}{41}{chapter.5}}
+\@writefile{toc}{\contentsline {chapter}{\numberline {5}Conclusion g\IeC {\'e}n\IeC {\'e}rale}{43}{chapter.5}}
 \@writefile{lof}{\addvspace {10\p@ }}
 \@writefile{lot}{\addvspace {10\p@ }}
 \bibcite{kddcup99}{1}
@@ -293,86 +310,95 @@
 \bibcite{aldinucci2012parallel}{5}
 \bibcite{arbelaez2011contour}{6}
 \bibcite{arora1998approximation}{7}
-\bibcite{bertaux2004speckle}{8}
-\bibcite{boykov2004experimental}{9}
-\bibcite{1467423}{10}
-\bibcite{BuadesCM06}{11}
-\bibcite{Caselles99topographicmaps}{12}
-\bibcite{cates2004gist}{13}
-\bibcite{chandran2009computational}{14}
-\bibcite{che2008performance}{15}
-\bibcite{chen09}{16}
-\bibcite{1093941}{17}
-\bibcite{cheng1995mean}{18}
-\bibcite{cherkassky1997implementing}{19}
-\bibcite{ChesnaudRB99}{20}
-\bibcite{cohen1993surface}{21}
-\bibcite{comaniciu1999mean}{22}
-\bibcite{comaniciu2002mean}{23}
-\bibcite{cutrona1990synthetic}{24}
-\bibcite{Dabov06imagedenoising}{25}
-\bibcite{Dabov09bm3dimage}{26}
-\bibcite{Daubechies:1992:TLW:130655}{27}
-\bibcite{dixit2005gpu}{28}
-\bibcite{elad2006image}{29}
-\bibcite{felzenszwalb2004efficient}{30}
-\bibcite{fluck2006gpu}{31}
-\bibcite{foley1994introduction}{32}
-\bibcite{ford1955simple}{33}
-\bibcite{fukunaga1975estimation}{34}
-\bibcite{fulkerson2012really}{35}
-\bibcite{GallandBR03}{36}
-\bibcite{GermainR01}{37}
-\bibcite{nlmeansgpubelge}{38}
-\bibcite{healey1994radiometric}{39}
-\bibcite{hochbaum2013simplifications}{40}
-\bibcite{5170921}{41}
-\bibcite{humphrey1924psychology}{42}
-\bibcite{jeong2009scalable}{43}
-\bibcite{5402362}{44}
-\bibcite{KassWT88}{45}
-\bibcite{keselman1998extraction}{46}
-\bibcite{cmla2009Kes}{47}
-\bibcite{kohli2007dynamic}{48}
-\bibcite{lefohn2003inter}{49}
-\bibcite{lefohn2003interactive}{50}
-\bibcite{lefohn2005streaming}{51}
-\bibcite{li2009mean}{52}
-\bibcite{macqueen1967some}{53}
-\bibcite{Mallat:2008:WTS:1525499}{54}
-\bibcite{mancuso2001introduction}{55}
-\bibcite{coil}{56}
-\bibcite{osher1988fronts}{57}
-\bibcite{4310076}{58}
-\bibcite{PALHANOXAVIERDEFONTES}{59}
-\bibcite{pelleg2000x}{60}
-\bibcite{4287006}{61}
-\bibcite{1521458}{62}
-\bibcite{4587843}{63}
-\bibcite{Roberts:2010:WGA:1921479.1921499}{64}
-\bibcite{ronfard1994region}{65}
-\bibcite{rumpf2001level}{66}
-\bibcite{6288187}{67}
-\bibcite{sethian1996fast}{68}
-\bibcite{kmeansgpuopengl}{69}
-\bibcite{shi2000normalized}{70}
-\bibcite{convolutionsoup}{71}
-\bibcite{graphcutscuda}{72}
-\bibcite{strang1999discrete}{73}
-\bibcite{theuwissen2001ccd}{74}
-\bibcite{710815}{75}
-\bibcite{tukey77}{76}
-\bibcite{vedaldi2008quick}{77}
-\bibcite{4563095}{78}
-\bibcite{Vineet:2009:FMS:1572769.1572796}{79}
-\bibcite{wang2001image}{80}
-\bibcite{wang2003image}{81}
-\bibcite{Wang04imagequality}{82}
-\bibcite{wu1993optimal}{83}
-\bibcite{xiao2010efficient}{84}
-\bibcite{5206542}{85}
-\bibcite{Zahn:1971:GMD:1309266.1309359}{86}
-\bibcite{zheng2011performance}{87}
+\bibcite{bauer2009segmentation}{8}
+\bibcite{bertaux2004speckle}{9}
+\bibcite{boykov2004experimental}{10}
+\bibcite{bresenham1965algorithm}{11}
+\bibcite{1467423}{12}
+\bibcite{BuadesCM06}{13}
+\bibcite{Caselles99topographicmaps}{14}
+\bibcite{5459410}{15}
+\bibcite{cates2004gist}{16}
+\bibcite{chandran2009computational}{17}
+\bibcite{che2008performance}{18}
+\bibcite{chen09}{19}
+\bibcite{1093941}{20}
+\bibcite{cheng1995mean}{21}
+\bibcite{cherkassky1997implementing}{22}
+\bibcite{ChesnaudRB99}{23}
+\bibcite{cohen1993surface}{24}
+\bibcite{comaniciu1999mean}{25}
+\bibcite{comaniciu2002mean}{26}
+\bibcite{cutrona1990synthetic}{27}
+\bibcite{Dabov06imagedenoising}{28}
+\bibcite{Dabov09bm3dimage}{29}
+\bibcite{Daubechies:1992:TLW:130655}{30}
+\bibcite{dixit2005gpu}{31}
+\bibcite{elad2006image}{32}
+\bibcite{felzenszwalb2004efficient}{33}
+\bibcite{fluck2006gpu}{34}
+\bibcite{foley1994introduction}{35}
+\bibcite{ford1955simple}{36}
+\bibcite{fukunaga1975estimation}{37}
+\bibcite{fulkerson2012really}{38}
+\bibcite{GallandBR03}{39}
+\bibcite{GermainR01}{40}
+\bibcite{nlmeansgpubelge}{41}
+\bibcite{snakegvf06}{42}
+\bibcite{healey1994radiometric}{43}
+\bibcite{hochbaum2013simplifications}{44}
+\bibcite{5170921}{45}
+\bibcite{humphrey1924psychology}{46}
+\bibcite{jeong2009scalable}{47}
+\bibcite{5402362}{48}
+\bibcite{KassWT88}{49}
+\bibcite{keselman1998extraction}{50}
+\bibcite{cmla2009Kes}{51}
+\bibcite{kohli2007dynamic}{52}
+\bibcite{lefohn2003inter}{53}
+\bibcite{lefohn2003interactive}{54}
+\bibcite{lefohn2005streaming}{55}
+\bibcite{li2009mean}{56}
+\bibcite{li2011robust}{57}
+\bibcite{macqueen1967some}{58}
+\bibcite{Mallat:2008:WTS:1525499}{59}
+\bibcite{mancuso2001introduction}{60}
+\bibcite{martin2001database}{61}
+\bibcite{coil}{62}
+\bibcite{osher1988fronts}{63}
+\bibcite{4310076}{64}
+\bibcite{PALHANOXAVIERDEFONTES}{65}
+\bibcite{pelleg2000x}{66}
+\bibcite{4287006}{67}
+\bibcite{1521458}{68}
+\bibcite{4587843}{69}
+\bibcite{Roberts:2010:WGA:1921479.1921499}{70}
+\bibcite{ronfard1994region}{71}
+\bibcite{rumpf2001level}{72}
+\bibcite{6288187}{73}
+\bibcite{sethian1996fast}{74}
+\bibcite{kmeansgpuopengl}{75}
+\bibcite{shi2000normalized}{76}
+\bibcite{snakegvfopencl12}{77}
+\bibcite{convolutionsoup}{78}
+\bibcite{graphcutscuda}{79}
+\bibcite{strang1999discrete}{80}
+\bibcite{theuwissen2001ccd}{81}
+\bibcite{710815}{82}
+\bibcite{tukey77}{83}
+\bibcite{vedaldi2008quick}{84}
+\bibcite{4563095}{85}
+\bibcite{Vineet:2009:FMS:1572769.1572796}{86}
+\bibcite{wang2001image}{87}
+\bibcite{wang2003image}{88}
+\bibcite{Wang04imagequality}{89}
+\bibcite{wu1993optimal}{90}
+\bibcite{xiao2010efficient}{91}
+\bibcite{5206542}{92}
+\bibcite{Zahn:1971:GMD:1309266.1309359}{93}
+\bibcite{zheng2011performance}{94}
+\bibcite{zheng2012fast}{95}
 \citation{zheng2011performance}
 \citation{graphcutscuda}
 \citation{fulkerson2012really}
+\citation{martin2001database}

diff --git a/THESE/these.bbl b/THESE/these.bbl
index 851ac749426b69084fe3b95f76e3b4f284ac9da7..1dd0d5a101f625ae4e6b0606dd3cfb470ede6d2a 100644 (file)
--- a/THESE/these.bbl
+++ b/THESE/these.bbl
@@ -35,6 +35,12 @@ Sanjeev Arora, Prabhakar Raghavan, and Satish Rao.
 \newblock In {\em Proceedings of the thirtieth annual ACM symposium on Theory
   of computing}, pages 106--113. ACM, 1998.
 
+\bibitem{bauer2009segmentation}
+Christian Bauer, Horst Bischof, and Reinhard Beichel.
+\newblock Segmentation of airways based on gradient vector flow.
+\newblock In {\em International workshop on pulmonary image analysis, Medical
+  image computing and computer assisted intervention}, pages 191--201, 2009.
+
 \bibitem{bertaux2004speckle}
 Nicolas Bertaux, Yann Frauel, Philippe R{\'e}fr{\'e}gier, and Bahram Javidi.
 \newblock Speckle removal using a maximum-likelihood technique with isoline
@@ -48,6 +54,11 @@ Yuri Boykov and Vladimir Kolmogorov.
 \newblock {\em Pattern Analysis and Machine Intelligence, IEEE Transactions
   on}, 26(9):1124--1137, 2004.
 
+\bibitem{bresenham1965algorithm}
+Jack~E Bresenham.
+\newblock Algorithm for computer control of a digital plotter.
+\newblock {\em IBM Systems journal}, 4(1):25--30, 1965.
+
 \bibitem{1467423}
 A.~Buades, B.~Coll, and J.~M Morel.
 \newblock A non-local algorithm for image denoising.
@@ -64,6 +75,13 @@ Vicent Caselles and Jean michel Morel.
 \newblock Topographic maps and local contrast changes in natural images.
 \newblock {\em Int. J. Comp. Vision}, 33:5--27, 1999.
 
+\bibitem{5459410}
+Bryan Catanzaro, Bor-Yiing Su, N.~Sundaram, Yunsup Lee, Mark Murphy, and
+  K.~Keutzer.
+\newblock Efficient, high-quality image contour detection.
+\newblock In {\em Computer Vision, 2009 IEEE 12th International Conference on},
+  pages 2381--2388, 2009.
+
 \bibitem{cates2004gist}
 Joshua~E Cates, Aaron~E Lefohn, and Ross~T Whitaker.
 \newblock Gist: an interactive, gpu-based level set segmentation tool for 3d
@@ -224,6 +242,15 @@ Bart Goossens, Hiêp Luong, Jan Aelterman, Aleksandra Pižurica, and Wilfried
   Systems}, volume 6475 of {\em Lecture Notes in Computer Science}, pages
   46--57. Springer Berlin Heidelberg, 2010.
 
+\bibitem{snakegvf06}
+Zhiyu He and Falko Kuester.
+\newblock Gpu-based active contour segmentation using gradient vector flow.
+\newblock In George Bebis, Richard Boyle, Bahram Parvin, Darko Koracin, Paolo
+  Remagnino, Ara Nefian, Gopi Meenakshisundaram, Valerio Pascucci, Jiri Zara,
+  Jose Molineros, Holger Theisel, and Tom Malzbender, editors, {\em Advances in
+  Visual Computing}, volume 4291 of {\em Lecture Notes in Computer Science},
+  pages 191--201. Springer Berlin Heidelberg, 2006.
+
 \bibitem{healey1994radiometric}
 Glenn~E Healey and Raghava Kondepudy.
 \newblock Radiometric ccd camera calibration and noise estimation.
@@ -309,6 +336,12 @@ Peihua Li and Lijuan Xiao.
 \newblock In {\em Pattern Recognition and Image Analysis}, pages 120--127.
   Springer, 2009.
 
+\bibitem{li2011robust}
+Tao Li, Alexandre Krupa, and Christophe Collewet.
+\newblock A robust parametric active contour based on fourier descriptors.
+\newblock In {\em Image Processing (ICIP), 2011 18th IEEE International
+  Conference on}, pages 1037--1040. IEEE, 2011.
+
 \bibitem{macqueen1967some}
 James MacQueen et~al.
 \newblock Some methods for classification and analysis of multivariate
@@ -327,6 +360,13 @@ Massimo Mancuso and Sebastiano Battiato.
 \newblock An introduction to the digital still camera technology.
 \newblock {\em ST Journal of System Research}, 2(2), 2001.
 
+\bibitem{martin2001database}
+David Martin, Charless Fowlkes, Doron Tal, and Jitendra Malik.
+\newblock A database of human segmented natural images and its application to
+  evaluating segmentation algorithms and measuring ecological statistics.
+\newblock In {\em Computer Vision, 2001. ICCV 2001. Proceedings. Eighth IEEE
+  International Conference on}, volume~2, pages 416--423. IEEE, 2001.
+
 \bibitem{coil}
 S.A. Nene, S.K. Nayar, and Murase H.
 \newblock Columbia object image library (coil-100).
@@ -418,6 +458,11 @@ Jianbo Shi and Jitendra Malik.
 \newblock {\em Pattern Analysis and Machine Intelligence, IEEE Transactions
   on}, 22(8):888--905, 2000.
 
+\bibitem{snakegvfopencl12}
+Erik Smistad, AnneC. Elster, and Frank Lindseth.
+\newblock Real-time gradient vector flow on gpus using opencl.
+\newblock {\em Journal of Real-Time Image Processing}, pages 1--8, 2012.
+
 \bibitem{convolutionsoup}
 J.~Stam.
 \newblock Convolution soup.
@@ -519,4 +564,9 @@ Z.~Zheng, W.~Xu, and K.~Mueller.
 \newblock {\em Workshop on High Performance Image Reconstruction (HPIR)}, pages
   52--55, 2011.
 
+\bibitem{zheng2012fast}
+Zuoyong Zheng and Ruixia Zhang.
+\newblock A fast gvf snake algorithm on the gpu.
+\newblock {\em image}, 2:4, 2012.
+
 \end{thebibliography}

diff --git a/THESE/these.blg b/THESE/these.blg
index c2ecb6553e699cd93e02e6c23b463ddc882f6587..db7083f2b904f196010b9540d62a55b0c6872e01 100644 (file)
--- a/THESE/these.blg
+++ b/THESE/these.blg
@@ -10,45 +10,45 @@ Warning--empty journal in cohen1993surface
 Warning--empty journal in dixit2005gpu
 Warning--can't use both volume and number fields in macqueen1967some
 Warning--can't use both volume and number fields in xiao2010efficient
-You've used 87 entries,
+You've used 95 entries,
             2118 wiz_defined-function locations,
-            996 strings with 18735 characters,
-and the built_in function-call counts, 32324 in all, are:
-= -- 3258
-> -- 1354
-< -- 42
-+ -- 549
-- -- 442
-* -- 2217
-:= -- 5073
-add.period$ -- 277
-call.type$ -- 87
-change.case$ -- 471
+            1037 strings with 20359 characters,
+and the built_in function-call counts, 36042 in all, are:
+= -- 3622
+> -- 1535
+< -- 50
++ -- 620
+- -- 502
+* -- 2473
+:= -- 5637
+add.period$ -- 304
+call.type$ -- 95
+change.case$ -- 519
 chr.to.int$ -- 0
-cite$ -- 95
-duplicate$ -- 1320
-empty$ -- 2626
-format.name$ -- 442
-if$ -- 7004
+cite$ -- 103
+duplicate$ -- 1479
+empty$ -- 2910
+format.name$ -- 502
+if$ -- 7833
 int.to.chr$ -- 0
-int.to.str$ -- 87
-missing$ -- 87
-newline$ -- 435
-num.names$ -- 174
-pop$ -- 525
+int.to.str$ -- 95
+missing$ -- 95
+newline$ -- 475
+num.names$ -- 192
+pop$ -- 582
 preamble$ -- 1
-purify$ -- 388
+purify$ -- 428
 quote$ -- 0
-skip$ -- 992
+skip$ -- 1122
 stack$ -- 0
-substring$ -- 2226
-swap$ -- 392
-text.length$ -- 42
+substring$ -- 2471
+swap$ -- 461
+text.length$ -- 50
 text.prefix$ -- 0
 top$ -- 0
-type$ -- 338
+type$ -- 370
 warning$ -- 8
-while$ -- 314
-width$ -- 89
-write$ -- 969
+while$ -- 349
+width$ -- 97
+write$ -- 1062
 (There were 8 warnings)

diff --git a/THESE/these.lof b/THESE/these.lof
index 894deccf6f0720c28cfa41f1b80e3b06e98a3680..5e7bab4b603472ea5b5db3722c932c0eb5d327cc 100644 (file)
--- a/THESE/these.lof
+++ b/THESE/these.lof
@@ -74,6 +74,7 @@
 \contentsline {figure}{\numberline {2.17}{\ignorespaces Segmentation d'images issues d'examens IRM par la m\IeC {\'e}thode des level set \IeC {\`a} bande \IeC {\'e}troite.}}{36}{figure.2.17}
 \contentsline {subfigure}{\numberline {(a)}{\ignorespaces {Cerveau 256$\times $256$\times $256 en 7~s}}}{36}{figure.2.17}
 \contentsline {subfigure}{\numberline {(b)}{\ignorespaces {Reins et aorte, 256$\times $256$\times $272 en 16~s}}}{36}{figure.2.17}
+\contentsline {figure}{\numberline {2.18}{\ignorespaces Extraction de contour par la version GPU de l'algorithme gPb. Les images sont issues de la base BSDS \cite {martin2001database}}}{38}{figure.2.18}
 \addvspace {10\p@ }
 \addvspace {10\p@ }
 \addvspace {10\p@ }

diff --git a/THESE/these.log b/THESE/these.log
index a5f5481d3edb4ddd28a8ac3bd1eda60539046a25..220bea1f98249e30e7855d78d9d92630967cdfd4 100644 (file)
--- a/THESE/these.log
+++ b/THESE/these.log
@@ -1,4 +1,4 @@
-This is pdfTeX, Version 3.1415926-1.40.10 (TeX Live 2009/Debian) (format=pdflatex 2012.12.6)  19 SEP 2013 15:27
+This is pdfTeX, Version 3.1415926-1.40.10 (TeX Live 2009/Debian) (format=pdflatex 2012.12.6)  20 SEP 2013 14:29
 entering extended mode
  %&-line parsing enabled.
 **these.tex
@@ -891,17 +891,14 @@ subscribe by emailing sympa@lirmm.fr with 'subscribe <list> <firstname name>'
 \algocf@algobox=\box77
 ) *** define extension value defensedate ****
 *** define extension value speciality ****
-*** define extension value universityname **** (./these.aux
-
-LaTeX Warning: Label `fig-meanshift-castle' multiply defined.
-
-)
+*** define extension value universityname **** (./these.aux)
 \openout1 = `these.aux'.
 
 LaTeX Font Info:    Checking defaults for OML/txmi/m/it on input line 74.
 LaTeX Font Info:    Try loading font information for OML+txmi on input line 74.
 
- (/usr/share/texmf-texlive/tex/latex/txfonts/omltxmi.fd
+
+(/usr/share/texmf-texlive/tex/latex/txfonts/omltxmi.fd
 File: omltxmi.fd 2000/12/15 v3.1
 )
 LaTeX Font Info:    ... okay on input line 74.
@@ -985,7 +982,7 @@ LaTeX Info: Redefining \Ref on input line 74.
 pdfTeX warning: pdflatex (file /usr/share/texmf/tex/latex/upmethodology-extensi
 ons/phd_thesis/spimufcphdthesis/spimufcphdthesis-frontpage.pdf): PDF inclusion:
  found PDF version <1.5>, but at most version <1.4> allowed
-<spimufcphdthesis-frontpage.pdf, id=159, 597.50829pt x 845.0471pt>
+<spimufcphdthesis-frontpage.pdf, id=163, 597.50829pt x 845.0471pt>
 File: spimufcphdthesis-frontpage.pdf Graphic file (type pdf)
 
 <use spimufcphdthesis-frontpage.pdf> [1
@@ -996,7 +993,7 @@ File: spimufcphdthesis-frontpage.pdf Graphic file (type pdf)
 e.pdf>] [2
 
 ]
-<spimufcphdthesis-p3-head.pdf, id=207, 598.235pt x 285.065pt>
+<spimufcphdthesis-p3-head.pdf, id=211, 598.235pt x 285.065pt>
 File: spimufcphdthesis-p3-head.pdf Graphic file (type pdf)
 
 <use spimufcphdthesis-p3-head.pdf>
@@ -1051,12 +1048,12 @@ Chapitre 2.
 (./Chapters/chapter2/chapter2.tex [11
 
 ] [12] [13]
-</home/zulu/Documents/these_gilles/THESE/codes/ny256.png, id=353, 256.96pt x 25
+</home/zulu/Documents/these_gilles/THESE/codes/ny256.png, id=358, 256.96pt x 25
 6.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/ny256.png Graphic file (typ
 e png)
  <use /home/zulu/Documents/these_gilles/THESE/codes/ny256.png>
-</home/zulu/Documents/these_gilles/THESE/codes/ny256_gauss25.png, id=354, 256.9
+</home/zulu/Documents/these_gilles/THESE/codes/ny256_gauss25.png, id=359, 256.9
 6pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/ny256_gauss25.png Graphic f
 ile (type png)
@@ -1072,7 +1069,7 @@ Underfull \hbox (badness 1997) in paragraph at lines 79--79
  []
 
 
-</home/zulu/Documents/these_gilles/THESE/codes/ny256_sap25.png, id=355, 256.96p
+</home/zulu/Documents/these_gilles/THESE/codes/ny256_sap25.png, id=360, 256.96p
 t x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/ny256_sap25.png Graphic fil
 e (type png)
@@ -1091,7 +1088,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 80--80
 /zulu/Documents/these_gilles/THESE/codes/ny256_gauss25.png (PNG copy)> </home/z
 ulu/Documents/these_gilles/THESE/codes/ny256_sap25.png (PNG copy)>]
 </home/zulu/Documents/these_gilles/THESE/codes/convo/ny256_gauss25_moy3.png, id
-=367, 256.96pt x 256.96pt>
+=372, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/convo/ny256_gauss25_moy3.pn
 g Graphic file (type png)
 
@@ -1103,7 +1100,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 101--101
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/convo/ny256_gauss25_moy5.png, id
-=368, 256.96pt x 256.96pt>
+=373, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/convo/ny256_gauss25_moy5.pn
 g Graphic file (type png)
 
@@ -1115,7 +1112,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 102--102
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/convo/ny256_gauss25_g3.png, id=3
-69, 256.96pt x 256.96pt>
+74, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/convo/ny256_gauss25_g3.png 
 Graphic file (type png)
 
@@ -1126,7 +1123,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 103--103
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/median/ny256_sap25_med3.png, id=
-373, 256.96pt x 256.96pt>
+378, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/median/ny256_sap25_med3.png
  Graphic file (type png)
 
@@ -1143,7 +1140,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 115--115
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/median/ny256_sap25_med3x2.png, i
-d=374, 256.96pt x 256.96pt>
+d=379, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/median/ny256_sap25_med3x2.p
 ng Graphic file (type png)
 
@@ -1160,7 +1157,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 116--116
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/median/ny256_sap25_med5.png, id=
-375, 256.96pt x 256.96pt>
+380, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/median/ny256_sap25_med5.png
  Graphic file (type png)
 
@@ -1180,7 +1177,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 117--117
 _moy5.png (PNG copy)> </home/zulu/Documents/these_gilles/THESE/codes/convo/ny25
 6_gauss25_g3.png (PNG copy)>]
 </home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_1_01.png,
- id=391, 256.96pt x 256.96pt>
+ id=396, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_1_01
 .png Graphic file (type png)
 
@@ -1193,7 +1190,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 132--132
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_1_05.png,
- id=392, 256.96pt x 256.96pt>
+ id=397, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_1_05
 .png Graphic file (type png)
 
@@ -1206,7 +1203,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 133--133
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_1_1.png, 
-id=393, 256.96pt x 256.96pt>
+id=398, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_1_1.
 png Graphic file (type png)
 
@@ -1219,7 +1216,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 134--134
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_2_01.png,
- id=394, 256.96pt x 256.96pt>
+ id=399, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_2_01
 .png Graphic file (type png)
 
@@ -1232,7 +1229,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 135--135
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_2_05.png,
- id=395, 256.96pt x 256.96pt>
+ id=400, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_2_05
 .png Graphic file (type png)
 
@@ -1245,7 +1242,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 136--136
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_2_1.png, 
-id=396, 256.96pt x 256.96pt>
+id=401, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_2_1.
 png Graphic file (type png)
 
@@ -1258,7 +1255,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 137--137
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_5_01.png,
- id=397, 256.96pt x 256.96pt>
+ id=402, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_5_01
 .png Graphic file (type png)
 
@@ -1271,7 +1268,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 138--138
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_5_05.png,
- id=398, 256.96pt x 256.96pt>
+ id=403, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_5_05
 .png Graphic file (type png)
 
@@ -1284,7 +1281,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 139--139
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_5_1.png, 
-id=399, 256.96pt x 256.96pt>
+id=404, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/bilat/ny_gauss25_bilat_5_1.
 png Graphic file (type png)
 
@@ -1313,27 +1310,27 @@ ny_gauss25_bilat_5_01.png (PNG copy)> </home/zulu/Documents/these_gilles/THESE/
 codes/bilat/ny_gauss25_bilat_5_05.png (PNG copy)> </home/zulu/Documents/these_g
 illes/THESE/codes/bilat/ny_gauss25_bilat_5_1.png (PNG copy)>]
 </home/zulu/Documents/these_gilles/THESE/codes/wave/ny256_gauss25_dwt20.png, id
-=436, 256.96pt x 256.96pt>
+=441, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/wave/ny256_gauss25_dwt20.pn
 g Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/codes/wave/ny256_gauss25_dwt20.png
 >
 </home/zulu/Documents/these_gilles/THESE/codes/wave/ny256_gauss25_dwt.png, id=4
-37, 256.96pt x 256.96pt>
+42, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/wave/ny256_gauss25_dwt.png 
 Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/codes/wave/ny256_gauss25_dwt.png>
 </home/zulu/Documents/these_gilles/THESE/codes/wave/ny256_gauss25_dwt70.png, id
-=438, 256.96pt x 256.96pt>
+=443, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/wave/ny256_gauss25_dwt70.pn
 g Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/codes/wave/ny256_gauss25_dwt70.png
 >
 </home/zulu/Documents/these_gilles/THESE/codes/nlmeans/ny256_gauss25_nlm_2_2_25
-.png, id=444, 256.96pt x 256.96pt>
+.png, id=449, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/nlmeans/ny256_gauss25_nlm_2
 _2_25.png Graphic file (type png)
 
@@ -1345,7 +1342,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 173--173
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/nlmeans/ny256_gauss25_nlm_2_5_25
-.png, id=445, 256.96pt x 256.96pt>
+.png, id=450, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/nlmeans/ny256_gauss25_nlm_2
 _5_25.png Graphic file (type png)
 
@@ -1357,7 +1354,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 174--174
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/nlmeans/ny256_gauss25_nlm_5_2_25
-.png, id=446, 256.96pt x 256.96pt>
+.png, id=451, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/nlmeans/ny256_gauss25_nlm_5
 _2_25.png Graphic file (type png)
 
@@ -1369,7 +1366,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 175--175
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/nlmeans/ny256_gauss25_nlm_5_5_25
-.png, id=447, 256.96pt x 256.96pt>
+.png, id=452, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/nlmeans/ny256_gauss25_nlm_5
 _5_25.png Graphic file (type png)
 
@@ -1381,7 +1378,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 176--176
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/bm3D/ny256_gauss25_bm3D.png, id=
-448, 256.96pt x 256.96pt>
+453, 256.96pt x 256.96pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/bm3D/ny256_gauss25_bm3D.png
  Graphic file (type png)
 
@@ -1414,14 +1411,14 @@ nput line 207.
 
 [20]
 </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/shmem_prefetch_z
-heng2011.png, id=496, 264.23718pt x 201.75375pt>
+heng2011.png, id=501, 264.23718pt x 201.75375pt>
 File: /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/shmem_prefe
 tch_zheng2011.png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/shmem_prefet
 ch_zheng2011.png> [21] [22 </home/zulu/Documents/these_gilles/THESE/Chapters/ch
 apter2/img/shmem_prefetch_zheng2011.png (PNG copy)>] [23]
-</home/zulu/Documents/these_gilles/THESE/codes/cochon256.png, id=531, 255.95625
+</home/zulu/Documents/these_gilles/THESE/codes/cochon256.png, id=536, 255.95625
 pt x 255.95625pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/cochon256.png Graphic file 
 (type png)
@@ -1438,14 +1435,14 @@ Underfull \hbox (badness 10000) in paragraph at lines 274--274
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/seg_histogramme/histo-cochon256.
-png, id=532, 289.883pt x 242.506pt>
+png, id=537, 289.883pt x 242.506pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/seg_histogramme/histo-cocho
 n256.png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/codes/seg_histogramme/histo-cochon
 256.png>
 </home/zulu/Documents/these_gilles/THESE/codes/seg_histogramme/cochon256-seghis
-to-101-255.png, id=533, 204.765pt x 204.765pt>
+to-101-255.png, id=538, 204.765pt x 204.765pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/seg_histogramme/cochon256-s
 eghisto-101-255.png Graphic file (type png)
 
@@ -1466,7 +1463,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 276--276
  []
 
 
-</home/zulu/Documents/these_gilles/THESE/codes/cochon256-sig25.png, id=534, 255
+</home/zulu/Documents/these_gilles/THESE/codes/cochon256-sig25.png, id=539, 255
 .95625pt x 255.95625pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/cochon256-sig25.png Graphic
  file (type png)
@@ -1478,14 +1475,14 @@ Underfull \hbox (badness 10000) in paragraph at lines 277--277
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/seg_histogramme/histo-cochon256-
-sig25.png, id=535, 287.474pt x 240.097pt>
+sig25.png, id=540, 287.474pt x 240.097pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/seg_histogramme/histo-cocho
 n256-sig25.png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/codes/seg_histogramme/histo-cochon
 256-sig25.png>
 </home/zulu/Documents/these_gilles/THESE/codes/seg_histogramme/cochon256-sig25-
-seghisto-99-255.png, id=536, 204.765pt x 204.765pt>
+seghisto-99-255.png, id=541, 204.765pt x 204.765pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/seg_histogramme/cochon256-s
 ig25-seghisto-99-255.png Graphic file (type png)
 
@@ -1521,28 +1518,28 @@ ESE/codes/seg_histogramme/histo-cochon256-sig25.png (PNG copy)> </home/zulu/Doc
 uments/these_gilles/THESE/codes/seg_histogramme/cochon256-sig25-seghisto-99-255
 .png (PNG copy)>] [25]
 </home/zulu/Documents/these_gilles/THESE/codes/graphe/cochon128_ncuts_2seg.png,
- id=573, 317.988pt x 251.339pt>
+ id=578, 317.988pt x 251.339pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/graphe/cochon128_ncuts_2seg
 .png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/codes/graphe/cochon128_ncuts_2seg.
 png>
 </home/zulu/Documents/these_gilles/THESE/codes/graphe/cochon128_ncuts_3seg.png,
- id=574, 317.988pt x 251.339pt>
+ id=579, 317.988pt x 251.339pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/graphe/cochon128_ncuts_3seg
 .png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/codes/graphe/cochon128_ncuts_3seg.
 png>
 </home/zulu/Documents/these_gilles/THESE/codes/graphe/cochon128_ncuts_4seg.png,
- id=575, 317.988pt x 251.339pt>
+ id=580, 317.988pt x 251.339pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/graphe/cochon128_ncuts_4seg
 .png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/codes/graphe/cochon128_ncuts_4seg.
 png>
 </home/zulu/Documents/these_gilles/THESE/codes/graphe/cochon128_ncuts_5seg.png,
- id=576, 317.988pt x 251.339pt>
+ id=581, 317.988pt x 251.339pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/graphe/cochon128_ncuts_5seg
 .png Graphic file (type png)
 
@@ -1553,56 +1550,56 @@ on128_ncuts_3seg.png (PNG copy)> </home/zulu/Documents/these_gilles/THESE/codes
 /graphe/cochon128_ncuts_4seg.png (PNG copy)> </home/zulu/Documents/these_gilles
 /THESE/codes/graphe/cochon128_ncuts_5seg.png (PNG copy)>]
 </home/zulu/Documents/these_gilles/THESE/codes/kmeans/cochon128_kmeans_2seg.png
-, id=606, 317.988pt x 251.339pt>
+, id=611, 317.988pt x 251.339pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/kmeans/cochon128_kmeans_2se
 g.png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/codes/kmeans/cochon128_kmeans_2seg
 .png>
 </home/zulu/Documents/these_gilles/THESE/codes/kmeans/cochon128_kmeans_3seg.png
-, id=607, 317.988pt x 251.339pt>
+, id=612, 317.988pt x 251.339pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/kmeans/cochon128_kmeans_3se
 g.png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/codes/kmeans/cochon128_kmeans_3seg
 .png>
 </home/zulu/Documents/these_gilles/THESE/codes/kmeans/cochon128_kmeans_4seg.png
-, id=608, 317.988pt x 251.339pt>
+, id=613, 317.988pt x 251.339pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/kmeans/cochon128_kmeans_4se
 g.png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/codes/kmeans/cochon128_kmeans_4seg
 .png>
 </home/zulu/Documents/these_gilles/THESE/codes/kmeans/cochon128_kmeans_5seg.png
-, id=609, 317.988pt x 251.339pt>
+, id=614, 317.988pt x 251.339pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/kmeans/cochon128_kmeans_5se
 g.png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/codes/kmeans/cochon128_kmeans_5seg
 .png>
 </home/zulu/Documents/these_gilles/THESE/codes/meanshift/cochon128_meanshift_r1
-00m100.png, id=617, 318.791pt x 251.339pt>
+00m100.png, id=622, 318.791pt x 251.339pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/meanshift/cochon128_meanshi
 ft_r100m100.png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/codes/meanshift/cochon128_meanshif
 t_r100m100.png>
 </home/zulu/Documents/these_gilles/THESE/codes/meanshift/cochon128_meanshift_r5
-0m100.png, id=618, 318.791pt x 251.339pt>
+0m100.png, id=623, 318.791pt x 251.339pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/meanshift/cochon128_meanshi
 ft_r50m100.png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/codes/meanshift/cochon128_meanshif
 t_r50m100.png>
 </home/zulu/Documents/these_gilles/THESE/codes/meanshift/cochon128_meanshift_r3
-5m100.png, id=619, 317.988pt x 251.339pt>
+5m100.png, id=624, 317.988pt x 251.339pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/meanshift/cochon128_meanshi
 ft_r35m100.png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/codes/meanshift/cochon128_meanshif
 t_r35m100.png>
 </home/zulu/Documents/these_gilles/THESE/codes/meanshift/cochon128_meanshift_r2
-5m100.png, id=620, 317.988pt x 251.339pt>
+5m100.png, id=625, 317.988pt x 251.339pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/meanshift/cochon128_meanshi
 ft_r25m100.png Graphic file (type png)
 
@@ -1622,7 +1619,7 @@ lles/THESE/codes/meanshift/cochon128_meanshift_r35m100.png (PNG copy)> </home/z
 ulu/Documents/these_gilles/THESE/codes/meanshift/cochon128_meanshift_r25m100.pn
 g (PNG copy)>]
 </home/zulu/Documents/these_gilles/THESE/codes/snake/cochon128_tradi_snake_it3.
-png, id=651, 138.116pt x 138.116pt>
+png, id=656, 138.116pt x 138.116pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/snake/cochon128_tradi_snake
 _it3.png Graphic file (type png)
 
@@ -1644,7 +1641,7 @@ Underfull \hbox (badness 2582) in paragraph at lines 409--409
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/snake/cochon128_tradi_snake_it7.
-png, id=652, 138.116pt x 138.116pt>
+png, id=657, 138.116pt x 138.116pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/snake/cochon128_tradi_snake
 _it7.png Graphic file (type png)
 
@@ -1656,7 +1653,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 410--410
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/snake/cochon128_tradi_snake_it10
-.png, id=653, 138.116pt x 138.116pt>
+.png, id=658, 138.116pt x 138.116pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/snake/cochon128_tradi_snake
 _it10.png Graphic file (type png)
 
@@ -1668,7 +1665,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 411--411
 
 
 </home/zulu/Documents/these_gilles/THESE/codes/snake/cochon128_tradi_snake_resu
-lt.png, id=654, 271.414pt x 271.414pt>
+lt.png, id=659, 271.414pt x 271.414pt>
 File: /home/zulu/Documents/these_gilles/THESE/codes/snake/cochon128_tradi_snake
 _result.png Graphic file (type png)
 
@@ -1698,7 +1695,7 @@ des/snake/cochon128_tradi_snake_it10.png (PNG copy)> </home/zulu/Documents/thes
 e_gilles/THESE/codes/snake/cochon128_tradi_snake_result.png (PNG copy)>]
 [31]
 </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/graphcutscuda_st
-itch.png, id=705, 552.56438pt x 232.61906pt>
+itch.png, id=710, 552.56438pt x 232.61906pt>
 File: /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/graphcutscu
 da_stitch.png Graphic file (type png)
 
@@ -1709,28 +1706,28 @@ Underfull \vbox (badness 1783) has occurred while \output is active []
  [33 </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/graphcutscu
 da_stitch.png (PNG copy)>]
 </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/quick-shift-yo-o
-rig.png, id=737, 262.73157pt x 263.48438pt>
+rig.png, id=742, 262.73157pt x 263.48438pt>
 File: /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/quick-shift
 -yo-orig.png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/quick-shift-
 yo-orig.png>
 </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/quick-shift-yo-s
-2t10.png, id=738, 263.48438pt x 262.73157pt>
+2t10.png, id=743, 263.48438pt x 262.73157pt>
 File: /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/quick-shift
 -yo-s2t10.png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/quick-shift-
 yo-s2t10.png>
 </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/quick-shift-yo-s
-10t10.png, id=739, 261.97874pt x 262.73157pt>
+10t10.png, id=744, 261.97874pt x 262.73157pt>
 File: /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/quick-shift
 -yo-s10t10.png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/quick-shift-
 yo-s10t10.png>
 </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/quick-shift-yo-s
-10t20.png, id=740, 261.22594pt x 263.48438pt>
+10t20.png, id=745, 261.22594pt x 263.48438pt>
 File: /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/quick-shift
 -yo-s10t20.png Graphic file (type png)
 
@@ -1743,14 +1740,14 @@ ge cou-leur de 512$\OMS/txsy/m/n/10.95 ^^B$\OT1/phv/m/n/10.95 512 pixels par
 
 
 </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/castle-meanshift
-.png, id=745, 267.24844pt x 355.3275pt>
+.png, id=750, 267.24844pt x 355.3275pt>
 File: /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/castle-mean
 shift.png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/castle-means
 hift.png>
 </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/castle-meanshift
--std.png, id=746, 265.74281pt x 355.3275pt>
+-std.png, id=751, 265.74281pt x 355.3275pt>
 File: /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/castle-mean
 shift-std.png Graphic file (type png)
 
@@ -1762,7 +1759,7 @@ Underfull \hbox (badness 2990) in paragraph at lines 501--501
 
 
 </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/castle-meanshift
--kdtree.png, id=747, 266.49562pt x 356.0803pt>
+-kdtree.png, id=752, 266.49562pt x 356.0803pt>
 File: /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/castle-mean
 shift-kdtree.png Graphic file (type png)
 
@@ -1785,14 +1782,14 @@ shift.png (PNG copy)> </home/zulu/Documents/these_gilles/THESE/Chapters/chapter
 2/img/castle-meanshift-std.png (PNG copy)> </home/zulu/Documents/these_gilles/T
 HESE/Chapters/chapter2/img/castle-meanshift-kdtree.png (PNG copy)>]
 </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/l7-brain7s.png, 
-id=788, 285.31593pt x 258.9675pt>
+id=792, 285.31593pt x 258.9675pt>
 File: /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/l7-brain7s.
 png Graphic file (type png)
 
 <use /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/l7-brain7s.p
 ng>
 </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/l7-reins16s.png,
- id=789, 264.99pt x 361.35pt>
+ id=793, 264.99pt x 361.35pt>
 File: /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/l7-reins16s
 .png Graphic file (type png)
 
@@ -1808,53 +1805,68 @@ Underfull \hbox (badness 10000) in paragraph at lines 519--519
 OT1/phv/m/n/9 272 en
  []
 
-) [36 </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/l7-brain7s
-.png (PNG copy)> </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img
-/l7-reins16s.png (PNG copy)>]
+
+LaTeX Warning: Reference `fig-snakegvf' on page 36 undefined on input line 528.
+
+
+
+Underfull \vbox (badness 10000) has occurred while \output is active []
+
+ [36 </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/l7-brain7s.
+png (PNG copy)> </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/
+l7-reins16s.png (PNG copy)>]
+</home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/gPb_examples.png
+, id=818, 655.69969pt x 175.40532pt>
+File: /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/gPb_example
+s.png Graphic file (type png)
+
+<use /home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/gPb_examples
+.png>) [37] [38 </home/zulu/Documents/these_gilles/THESE/Chapters/chapter2/img/
+gPb_examples.png (PNG copy)>]
 Chapitre 3.
-[37
+[39
 
-] [38
+] [40
 
 ]
 Chapitre 4.
-[39] [40
+[41] [42
 
 ]
 Chapitre 5.
-(./these.bbl [41] [42
+(./these.bbl [43] [44
 
-] [43] [44] [45] [46] [47]) [48] (./these.lof
+] [45] [46] [47] [48] [49]) [50] (./these.lof
 Underfull \vbox (badness 1448) has occurred while \output is active []
 
- [49
+ [51
 
 ])
 \tf@lof=\write5
 \openout5 = `these.lof'.
 
 
-[50] (./these.lot)
+[52] (./these.lot)
 \tf@lot=\write6
 \openout6 = `these.lot'.
 
- [51
+ [53
 
-] [52
+] [54
 
 
-] [53]
+] [55]
 
 pdfTeX warning: pdflatex (file /usr/share/texmf/tex/latex/upmethodology-extensi
 ons/phd_thesis/spimufcphdthesis/spimufcphdthesis-backpage.pdf): PDF inclusion: 
 found PDF version <1.6>, but at most version <1.4> allowed
-<spimufcphdthesis-backpage.pdf, id=889, 597.432pt x 844.83629pt>
+<spimufcphdthesis-backpage.pdf, id=930, 597.432pt x 844.83629pt>
 File: spimufcphdthesis-backpage.pdf Graphic file (type pdf)
 
 <use spimufcphdthesis-backpage.pdf>
 LaTeX Font Info:    Font shape `OT1/phv/bx/n' in size <12> not available
 (Font)              Font shape `OT1/phv/b/n' tried instead on input line 127.
- [54
+ [56
 
  </usr/share/texmf/tex/latex/upmethodology-extensions/phd_thesis/spimufcphdthes
 is/spimufcphdthesis-backpage.pdf>]
@@ -1862,16 +1874,13 @@ is/spimufcphdthesis-backpage.pdf>]
 
 LaTeX Warning: There were undefined references.
 
-
-LaTeX Warning: There were multiply-defined labels.
-
  )
 (\end occurred when \iftrue on line 127 was incomplete) 
 Here is how much of TeX's memory you used:
- 10473 strings out of 495028
- 164691 string characters out of 1181229
- 276122 words of memory out of 3000000
- 13018 multiletter control sequences out of 15000+50000
+ 10502 strings out of 495028
+ 165429 string characters out of 1181229
+ 276621 words of memory out of 3000000
+ 13034 multiletter control sequences out of 15000+50000
  71443 words of font info for 129 fonts, out of 3000000 for 9000
  28 hyphenation exceptions out of 8191
  61i,15n,47p,1405b,525s stack positions out of 5000i,500n,10000p,200000b,50000s
@@ -1886,9 +1895,9 @@ hare/texmf-texlive/fonts/type1/urw/helvetic/uhvr8a.pfb></usr/share/texmf-texliv
 e/fonts/type1/urw/helvetic/uhvro8a.pfb></usr/share/texmf-texlive/fonts/type1/ur
 w/times/utmr8a.pfb></usr/share/texmf-texlive/fonts/type1/urw/times/utmri8a.pfb>
 
-Output written on these.pdf (54 pages, 5946548 bytes).
+Output written on these.pdf (56 pages, 6228140 bytes).
 PDF statistics:
- 995 PDF objects out of 1000 (max. 8388607)
- 269 named destinations out of 1000 (max. 500000)
- 623 words of extra memory for PDF output out of 10000 (max. 10000000)
+ 1038 PDF objects out of 1200 (max. 8388607)
+ 281 named destinations out of 1000 (max. 500000)
+ 636 words of extra memory for PDF output out of 10000 (max. 10000000)
 

diff --git a/THESE/these.out b/THESE/these.out
index 666a7ec09e9b00184f10558882e9de60aba13ff3..69afe0e88aa22308f46e9e041d145aa705f964ed 100644 (file)
--- a/THESE/these.out
+++ b/THESE/these.out
@@ -29,6 +29,7 @@
 \BOOKMARK [2][]{subsection.2.6.2}{2.6.2 Partitionnement de graphe}{section.2.6}
 \BOOKMARK [2][]{subsection.2.6.3}{2.6.3 K-means, mean-shift et apparent\351s}{section.2.6}
 \BOOKMARK [2][]{subsection.2.6.4}{2.6.4 Snakes et Level set}{section.2.6}
+\BOOKMARK [2][]{subsection.2.6.5}{2.6.5 Algorithmes hybrides}{section.2.6}
 \BOOKMARK [0][]{chapter.3}{3 La segmentation orient\351e r\351gions dans les images bruit\351es}{}
 \BOOKMARK [1][]{section.3.1}{3.1 Pr\351sentation - existant}{chapter.3}
 \BOOKMARK [1][]{section.3.2}{3.2 La parall\350lisation du snake polygonal}{chapter.3}

diff --git a/THESE/these.pdf b/THESE/these.pdf
index 7180bd855890162fd03713e222648f01f12c98f6..212ecd6e563e6815584e85a46cdf712e5752ec3a 100644 (file)
Binary files a/THESE/these.pdf and b/THESE/these.pdf differ

diff --git a/THESE/these.tex b/THESE/these.tex
index a2a17e3200487092f469073ba99faa528229cd97..c0b8ada819f4d7d53e35139ca8f8c706043d8462 100644 (file)
--- a/THESE/these.tex
+++ b/THESE/these.tex
@@ -1,6 +1,6 @@
 \documentclass[french]{spimufcphdthesis}
 \usepackage[utf8]{inputenc}
-          
+           
 \usepackage{graphicx}
 \usepackage{color, xcolor}
 \usepackage[ruled,lined,linesnumbered]{algorithm2e}

diff --git a/THESE/these.toc b/THESE/these.toc
index 8752a7809b0568fc92d87a2cdee809cbe6b030bd..96de7f57eeb2ba736281fd2e12165a91f4300d44 100644 (file)
--- a/THESE/these.toc
+++ b/THESE/these.toc
@@ -30,11 +30,12 @@
 \contentsline {subsection}{\numberline {2.6.2}Partitionnement de graphe}{32}{subsection.2.6.2}
 \contentsline {subsection}{\numberline {2.6.3}K-means, mean-shift et apparent\IeC {\'e}s}{32}{subsection.2.6.3}
 \contentsline {subsection}{\numberline {2.6.4}Snakes et Level set}{34}{subsection.2.6.4}
-\contentsline {chapter}{\numberline {3}La segmentation orient\IeC {\'e}e r\IeC {\'e}gions dans les images bruit\IeC {\'e}es}{37}{chapter.3}
-\contentsline {section}{\numberline {3.1}Pr\IeC {\'e}sentation - existant}{37}{section.3.1}
-\contentsline {section}{\numberline {3.2}La parall\IeC {\`e}lisation du snake polygonal}{37}{section.3.2}
-\contentsline {chapter}{\numberline {4}Le filtrage des images sur GPU}{39}{chapter.4}
-\contentsline {section}{\numberline {4.1}Algorithme de r\IeC {\'e}duction de bruit par recherche des lignes de niveaux}{39}{section.4.1}
-\contentsline {section}{\numberline {4.2}Filtre m\IeC {\'e}dian}{39}{section.4.2}
-\contentsline {section}{\numberline {4.3}Filtres de convolution}{39}{section.4.3}
-\contentsline {chapter}{\numberline {5}Conclusion g\IeC {\'e}n\IeC {\'e}rale}{41}{chapter.5}
+\contentsline {subsection}{\numberline {2.6.5}Algorithmes hybrides}{37}{subsection.2.6.5}
+\contentsline {chapter}{\numberline {3}La segmentation orient\IeC {\'e}e r\IeC {\'e}gions dans les images bruit\IeC {\'e}es}{39}{chapter.3}
+\contentsline {section}{\numberline {3.1}Pr\IeC {\'e}sentation - existant}{39}{section.3.1}
+\contentsline {section}{\numberline {3.2}La parall\IeC {\`e}lisation du snake polygonal}{39}{section.3.2}
+\contentsline {chapter}{\numberline {4}Le filtrage des images sur GPU}{41}{chapter.4}
+\contentsline {section}{\numberline {4.1}Algorithme de r\IeC {\'e}duction de bruit par recherche des lignes de niveaux}{41}{section.4.1}
+\contentsline {section}{\numberline {4.2}Filtre m\IeC {\'e}dian}{41}{section.4.2}
+\contentsline {section}{\numberline {4.3}Filtres de convolution}{41}{section.4.3}
+\contentsline {chapter}{\numberline {5}Conclusion g\IeC {\'e}n\IeC {\'e}rale}{43}{chapter.5}