Le but du rendu stylisé est de produire un rendud'une scène 3D dans le style visuel particuliervoulu par un artiste.Cela nécessite de reproduire automatiquementsur ordinateur certaines caractéristiquesd'illustrations traditionnelles: par exemple,la façon dont un artiste représente les ombres et lalumière, les contours des objets, ou bien les coupsde pinceau qui ont servi à créer une peinture.Les problématiques du rendu stylisé sont pertinentesdans des domaines comme la réalisation de films d'animation 3Dou le jeu vidéo, où les studios cherchent de plus en plus à se démarquerpar des styles visuels originaux.Dans cette thèse, nous explorons des techniques destylisation qui peuvent s'intégrer dans des pipelinesde rendu temps-réel existants, et nous proposons deux contributions.La première est un outil de création de modèles d'illuminationstylisés pour des objets 3D.La conception de ces modèles est complexe et coûteuse en temps,car ils doivent produire un résultat cohérentsous une multitude d'angles de vue et d'éclairages.Nous proposons une méthode qui facilite la créationde modèles d'illumination pour le rendu stylisé,en les décomposant en sous-modèles plus simples à manipuler.Notre seconde contribution est un pipeline de rendude scènes 3D dans un style peinture,qui utilise une combinaison de bruits procéduraux 3Det de filtrage en espace écran.Des techniques de filtrage d'image ont déjà été proposéespour styliser des images ou des vidéos:le but de ce travail est d'utiliser ces filtres pourstyliser des scènes 3D tout en gardant la cohérence du mouvement.Cependant, directement appliquer un filtreen espace écran produit des défauts visuels au niveau dessilhouettes des objets.Nous proposons une méthode qui permet d'assurer la cohérence du mouvement,en guidant les filtres d'images avec des informations sur la géométrie extraites de G-buffers, et qui élimine les défauts aux silhouettes. / The goal of stylized rendering is to render 3D scenes in the visual style intended by an artist.This often entails reproducing, with some degree of automation,the visual features typically found in 2D illustrationsthat constitute the "style" of an artist.Examples of these features include the depiction of light and shade,the representation of the contours of objects,or the strokes on a canvas that make a painting.This field is relevant today in domains such as computer-generated animation orvideo games, where studios seek to differentiate themselveswith styles that deviate from photorealism.In this thesis, we explore stylization techniques that can be easilyinserted into existing real-time rendering pipelines, and propose two novel techniques in this domain.Our first contribution is a workflow that aims to facilitatethe design of complex stylized shading models for 3D objects.Designing a stylized shading model that follows artistic constraintsand stays consistent under a variety of lightingconditions and viewpoints is a difficult and time-consuming process.Specialized shading models intended for stylization existbut are still limited in the range of appearances and behaviors they can reproduce.We propose a way to build and experiment with complex shading modelsby combining several simple shading behaviors using a layered approach,which allows a more intuitive and efficient exploration of the design space of shading models.In our second contribution, we present a pipeline to render 3D scenes in painterly styles,simulating the appearance of brush strokes,using a combination of procedural noise andlocal image filtering in screen-space.Image filtering techniques can achieve a wide range of stylized effects on 2D pictures and video:our goal is to use those existing filtering techniques to stylize 3D scenes,in a way that is coherent with the underlying animation or camera movement.This is not a trivial process, as naive approaches to filtering in screen-spacecan introduce visual inconsistencies around the silhouette of objects.The proposed method ensures motion coherence by guiding filters with informationfrom G-buffers, and ensures a coherent stylization of silhouettes in a generic way.

This Master s thesis proposes a non-assisted procedural method for 3D canyons scenes generation based on techniques of computer graphics, computer vision and graph search algorithm. In order to define all the features to be reproduced in our scenes, we have analyzed several images of real canyons and have categorized them in two canyon features models: a recursive and an ordinary one. The proposed approach manipulates a heightmap, created using Perlin noise, in order to imitate the geological features formation previously analyzed. Several parametrizations are used to guide and constraint the generation of terrains, canyons features, course of river, plain areas, soft slope regions, cliffs and plateaus. This work also uses the Mean Shift algorithm as mechanism of segmentation to define regions of interest. A binary mask, with plain areas, is defined based on a threshold operation by a given data set provided by the Mean Shift algorithm. Thereafter a connected-component labeling algorithm is executed using the previously binary mask. This algorithm finds all plains centroids. Right after that, the Dijkstra s algorithm is performed in order to connect all plain areas, creating a valid path between the centroids. The Dijkstra s algorithm is executed again to define the river s course. Finally, a Gaussian smoothing operation is applied to interpolate the soft slope regions. The combination of all those techniques produces as a result automatically generated feature-rich canyons. / Esta dissertação propõe um método procedural não assistido, baseado em técnicas de computação gráfica, visão computacional e busca em grafos, para a geração de cenários 3D de cânions com foco em jogos digitais. Para definir as características a serem reproduzidas, foram analisadas diversas imagens de cânions reais chegando-se em dois modelos, um comum e outro recursivo. A abordagem proposta manipula um reticulado gerado com ruído de Perlin, moldando assim as características inerentes a essa formação geológica. São levadas em conta as diversas parametrizações necessárias para permitir que o algoritmo construa cânions com curso de rio, áreas de planícies, regiões de encosta suave, estruturas de penhascos e, por fim, planaltos nas regiões mais altas. Para atingir o resultado final, o trabalho utiliza o algoritmo Mean Shift como mecanismo de segmentação, definindo dados e regiões de interesse. Munido dos dados do algoritmo de clusterizacao, é definido um limiar para a criação de uma máscara binária com a definição das planícies. Em um segundo momento, um algoritmo de rotulação de componentes conectados é executado, extraindo-se os centróides de cada planície. Por sua vez, o algoritmo de Dijkstra encaixa-se na definição de rotas que conectam estas planícies. O algoritmo de Dijkstra é, então, executado novamente, tendo por base uma função de custo de inclinação, para definir o curso do rio. Por fim, uma filtragem espacial baseada em um filtro Gaussiano é aplicada para interpolar as regiões de encostas de declive suave. A combinação dessas técnicas gera terrenos com grande variabilidade e com as características inerentes à formação geológica de cânions.