\select@language {french}
\contentsline {chapter}{\numberline {1}Introduction}{9}{chapter.1}
\contentsline {chapter}{\numberline {2}Les processeurs graphiques (GPU) NVidia\textregistered }{13}{chapter.2}
\contentsline {section}{\numberline {2.1}Pourquoi ?}{13}{section.2.1}
\contentsline {section}{\numberline {2.2}Comment ?}{14}{section.2.2}
\contentsline {subsection}{\numberline {2.2.1}Le mat\IeC {\'e}riel}{14}{subsection.2.2.1}
\contentsline {subsection}{\numberline {2.2.2}Le logiciel}{16}{subsection.2.2.2}
\contentsline {subsection}{\numberline {2.2.3}L'occupancy}{17}{subsection.2.2.3}
\contentsline {chapter}{\numberline {3}Mod\IeC {\`e}les d'image et de bruits - notations}{23}{chapter.3}
\contentsline {section}{\numberline {3.1}Mod\IeC {\`e}le d'image bruit\IeC {\'e}e}{23}{section.3.1}
\contentsline {section}{\numberline {3.2}Mod\IeC {\`e}les de bruit}{23}{section.3.2}
\contentsline {subsection}{\numberline {3.2.1}Le bruit gaussien}{23}{subsection.3.2.1}
\contentsline {subsection}{\numberline {3.2.2}Le speckle}{24}{subsection.3.2.2}
\contentsline {subsection}{\numberline {3.2.3}Le bruit \FB@og sel et poivre \FB@fg {}}{24}{subsection.3.2.3}
\contentsline {subsection}{\numberline {3.2.4}Le bruit de Poisson}{25}{subsection.3.2.4}
\contentsline {chapter}{\numberline {4}Les techniques de r\IeC {\'e}duction de bruit}{27}{chapter.4}
\contentsline {section}{\numberline {4.1}Les techniques de r\IeC {\'e}duction de bruit}{27}{section.4.1}
\contentsline {subsection}{\numberline {4.1.1}Les op\IeC {\'e}rateurs de base}{28}{subsection.4.1.1}
\contentsline {subsubsection}{\numberline {4.1.1.1}Le filtre de convolution}{28}{subsubsection.4.1.1.1}
\contentsline {subsubsection}{\numberline {4.1.1.2}Le filtre m\IeC {\'e}dian}{29}{subsubsection.4.1.1.2}
\contentsline {subsubsection}{\numberline {4.1.1.3}Le filtre bilat\IeC {\'e}ral}{30}{subsubsection.4.1.1.3}
\contentsline {subsubsection}{\numberline {4.1.1.4}Les algorithmes de filtrage par dictionnaire}{30}{subsubsection.4.1.1.4}
\contentsline {subsection}{\numberline {4.1.2}Les algorithmes de filtrage par patches}{32}{subsection.4.1.2}
\contentsline {section}{\numberline {4.2}Les impl\IeC {\'e}mentations sur GPU des algorithmes de filtrage}{33}{section.4.2}
\contentsline {subsection}{\numberline {4.2.1}Le filtrage par convolution}{33}{subsection.4.2.1}
\contentsline {subsection}{\numberline {4.2.2}Le filtre m\IeC {\'e}dian}{34}{subsection.4.2.2}
\contentsline {subsection}{\numberline {4.2.3}Le filtre bilat\IeC {\'e}ral}{35}{subsection.4.2.3}
\contentsline {subsection}{\numberline {4.2.4}Les filtres par patches}{37}{subsection.4.2.4}
\contentsline {chapter}{\numberline {5}Les techniques de segmentation des images}{39}{chapter.5}
\contentsline {section}{\numberline {5.1}Les techniques de segmentation}{39}{section.5.1}
\contentsline {subsection}{\numberline {5.1.1}Analyse d'histogramme}{40}{subsection.5.1.1}
\contentsline {subsection}{\numberline {5.1.2}Partitionnement de graphe}{40}{subsection.5.1.2}
\contentsline {subsection}{\numberline {5.1.3}kernel-means, mean-shift et apparent\IeC {\'e}s}{43}{subsection.5.1.3}
\contentsline {subsection}{\numberline {5.1.4}Les contours actifs, ou snakes}{45}{subsection.5.1.4}
\contentsline {subsection}{\numberline {5.1.5}M\IeC {\'e}thodes hybrides}{47}{subsection.5.1.5}
\contentsline {section}{\numberline {5.2}Les impl\IeC {\'e}mentations des techniques de segmentation sur GPU}{47}{section.5.2}
\contentsline {subsection}{\numberline {5.2.1}Calcul d'histogramme}{48}{subsection.5.2.1}
\contentsline {subsection}{\numberline {5.2.2}Partitionnement de graphe}{48}{subsection.5.2.2}
\contentsline {subsection}{\numberline {5.2.3}K-means, mean-shift et apparent\IeC {\'e}s}{49}{subsection.5.2.3}
\contentsline {subsection}{\numberline {5.2.4}Level set et snakes}{52}{subsection.5.2.4}
\contentsline {subsection}{\numberline {5.2.5}Algorithmes hybrides}{54}{subsection.5.2.5}
\contentsline {section}{\numberline {5.3}Conclusion}{55}{section.5.3}
\contentsline {chapter}{\numberline {6}La segmentation par snake polygonal orient\IeC {\'e} r\IeC {\'e}gions}{59}{chapter.6}
\contentsline {section}{\numberline {6.1}Introduction}{59}{section.6.1}
\contentsline {section}{\numberline {6.2}Pr\IeC {\'e}sentation de l'algorithme}{59}{section.6.2}
\contentsline {subsection}{\numberline {6.2.1}Formulation}{59}{subsection.6.2.1}
\contentsline {subsection}{\numberline {6.2.2}Optimisation des calculs}{60}{subsection.6.2.2}
\contentsline {subsection}{\numberline {6.2.3}Impl\IeC {\'e}mentation s\IeC {\'e}quentielle}{62}{subsection.6.2.3}
\contentsline {subsection}{\numberline {6.2.4}Performances}{65}{subsection.6.2.4}
\contentsline {section}{\numberline {6.3}Impl\IeC {\'e}mentation parall\IeC {\`e}le GPU du snake polygonal}{68}{section.6.3}
\contentsline {subsection}{\numberline {6.3.1}Pr\IeC {\'e}-calculs des images cumul\IeC {\'e}es}{69}{subsection.6.3.1}
\contentsline {subsection}{\numberline {6.3.2}Calcul des contributions des segments}{71}{subsection.6.3.2}
\contentsline {subsubsection}{\numberline {6.3.2.1}Cas particulier des segments dont la pente $k$ v\IeC {\'e}rifie $|k|\leq 1$}{74}{subsubsection.6.3.2.1}
\contentsline {subsection}{\numberline {6.3.3}Performances}{74}{subsection.6.3.3}
\contentsline {subsection}{\numberline {6.3.4}D\IeC {\'e}termination intelligente du contour initial}{76}{subsection.6.3.4}
\contentsline {subsection}{\numberline {6.3.5}Conclusion}{77}{subsection.6.3.5}
\contentsline {chapter}{\numberline {7}R\IeC {\'e}duction de bruit par recherche des lignes de niveaux}{81}{chapter.7}
\contentsline {section}{\numberline {7.1}Introduction}{81}{section.7.1}
\contentsline {section}{\numberline {7.2}Pr\IeC {\'e}sentation de l'algorithme}{82}{section.7.2}
\contentsline {subsection}{\numberline {7.2.1}Formulation}{82}{subsection.7.2.1}
\contentsline {subsubsection}{\numberline {7.2.1.1}Isolines \IeC {\`a} un seul segment}{82}{subsubsection.7.2.1.1}
\contentsline {subsubsection}{\numberline {7.2.1.2}Isolines compos\IeC {\'e}es de plusieurs segments - crit\IeC {\`e}re d'allongement}{84}{subsubsection.7.2.1.2}
\contentsline {section}{\numberline {7.3}Mod\IeC {\'e}lisation des isolines pour l'impl\IeC {\'e}mentation parall\IeC {\`e}le sur GPU}{85}{section.7.3}
\contentsline {subsection}{\numberline {7.3.1}Isolines \IeC {\'e}valu\IeC {\'e}es semi-globalement}{86}{subsection.7.3.1}
\contentsline {subsection}{\numberline {7.3.2}Isolines \IeC {\`a} segments pr\IeC {\'e}-\IeC {\'e}valu\IeC {\'e}s - mod\IeC {\`e}le PI-PD}{88}{subsection.7.3.2}
\contentsline {subsection}{\numberline {7.3.3}Mod\IeC {\`e}le PI-PD hybride}{90}{subsection.7.3.3}
\contentsline {subsubsection}{\numberline {7.3.3.1}Le d\IeC {\'e}tecteur de zone \IeC {\`a} faible pente}{91}{subsubsection.7.3.3.1}
\contentsline {section}{\numberline {7.4}R\IeC {\'e}sultats}{94}{section.7.4}
\contentsline {section}{\numberline {7.5}Extension aux images couleurs}{96}{section.7.5}
\contentsline {subsection}{\numberline {7.5.1}Expression du crit\IeC {\`e}re}{96}{subsection.7.5.1}
\contentsline {subsection}{\numberline {7.5.2}R\IeC {\'e}sultats}{97}{subsection.7.5.2}
\contentsline {section}{\numberline {7.6}Conclusion}{99}{section.7.6}
\contentsline {chapter}{\numberline {8}Le filtre m\IeC {\'e}dian sur GPU}{103}{chapter.8}
\contentsline {section}{\numberline {8.1}Introduction}{103}{section.8.1}
\contentsline {section}{\numberline {8.2}Les transferts de donn\IeC {\'e}es}{103}{section.8.2}
\contentsline {section}{\numberline {8.3}Utilisation des registres}{104}{section.8.3}
\contentsline {subsection}{\numberline {8.3.1}La s\IeC {\'e}lection de la valeur m\IeC {\'e}diane}{105}{subsection.8.3.1}
\contentsline {subsection}{\numberline {8.3.2}Masquage des latences}{107}{subsection.8.3.2}
\contentsline {section}{\numberline {8.4}R\IeC {\'e}sultats}{109}{section.8.4}
\contentsline {section}{\numberline {8.5}Conclusion}{111}{section.8.5}
\contentsline {chapter}{\numberline {9}Les filtres de convolution sur GPU}{113}{chapter.9}
\contentsline {section}{\numberline {9.1}Introduction}{113}{section.9.1}
\contentsline {section}{\numberline {9.2}Impl\IeC {\'e}mentation g\IeC {\'e}n\IeC {\'e}rique de la convolution non s\IeC {\'e}parable sur GPU}{113}{section.9.2}
\contentsline {section}{\numberline {9.3}Impl\IeC {\'e}mentation optimis\IeC {\'e}e de la convolution non s\IeC {\'e}parable sur GPU}{115}{section.9.3}
\contentsline {section}{\numberline {9.4}Cas de la convolution s\IeC {\'e}parable}{119}{section.9.4}
\contentsline {section}{\numberline {9.5}Conclusion}{124}{section.9.5}
\contentsline {chapter}{\numberline {10}Conclusion g\IeC {\'e}n\IeC {\'e}rale}{127}{chapter.10}