L'objectif de cette thèse est lerendu de scène virtuelles extrêmement détaillées.Nous nous intéressons plus particulièrementaux algorithmes de rendu de haute qualitéreposant sur du path tracing qui sont très largement utilisésdans l'industrie des effets spéciaux et pour calculerle rendu des films d'animations.Les grandes quantités de détail nécessaires à la modélisationde mondes virtuels crédibles soulèvent de sérieux problèmes d'efficacitédu rendu qui paralysent les studios et compliquent grandementle travail des artistes. Nous introduisons de nouveaux algorithmespour préfiltrer les objets 3D complexes à des échelles arbitraires,afin de réduire les temps de chargement, les coûts d'intersection des rayons lumineux,le calcul des matériaux et la quantité de bruit dans les images,le tout sans porter atteinte à la qualité du rendu.Nos contributions principales sont une nouvelle approchehybride de niveaux de détail qui allie les avantages des maillageset des représentations volumiques pour le préfiltrage des objets complexes àdes échelles arbitraires, ainsi qu'une nouvelle approchede préfiltrage pour le cas des volumes hétérogènes de haute résolution. / We address the problem of rendering extremely complex virtualscenes with large amounts of detail. We focuson high-end off-line rendering algorithms based on path tracing thatare intensively used in the special effects and 3D animation industries.The large amounts of detail required for creating believable virtual worldsraise important efficiency problemsthat paralyze production rendering pipelines andgreatly complicate the work of 3D artists.We introduce new algorithms for prefilteringcomplex 3D assets while preserving their appearanceat all scales, in order to reduce loading times,ray intersection costs, shading costs and Monte Carlo noise,without lowering the quality of rendered frames.Our main contributions are a new hybrid LODapproach that combines the benefits ofmeshes and volumetric representationsfor prefiltering complex 3D assets,as well as a new approach for prefilteringhigh-resolution heterogeneous participatingmedia.

Background To be able to push game standards further and further, different graphics techniques have to be implemented by game developers. By utilizing some of those techniques, comes cost in the form of performance issues. Level of detail (LOD) is an optimization technique which uses several different variations of a model and distance from the viewer to determine which model should be rendered. Shadow mapping is, on the contrary from level of detail, a costly technique which aims to create as realistically looking shadows as possible. Objectives The aim of this thesis is to examine whether a novel LOD technique and shadow map technique called LOD-SM can achieve less render time versus using a standard implementation of shadow mapping. This is achieved by using the Unity game engine in which the LOD-SM technique is implemented. With Unity, measurements will be taken to determine if the technique is a viable optimization option in future game development. Methods The methods used to answer the paper’s research question were a pilot user study and a performance study. 20 participants were part of the pilot study, and were asked to rate their perception of shadows shown on a screen in front of them. This yielded results of which shadows the participants noticed differences for. The results were then implemented into Unity and measured with the LOD-SM technique. Results The pilot study showed that user does not notice any differences in 30% decreased polycount-shadows when in a multiple building environment. The gathered results from the performance experiment showed that the LOD-SM technique saved between ∼17% up to ∼70% in render time, compared to a regular implementation of shadow mapping. Conclusions In conclusion, a viable optimization option was found which is to lower a shadow maps appearance based on buildings LODs because as the results has shown, users does not notice the lowering of LODs up to 30% when in a multiple building area. It was discovered that by measuring the LOD-SM technique, clear evidence of optimization was found when using the technique versus using a normal implementation of shadow mapping. Developers would have to consider the size of project and its models before implementing the LOD-SM technique, as it might not be viable for projects with a total polycount of less than 100 000 polygons.

Background. A previous study on performance benefits of joint reduction for animations was made by the authors of this thesis. The results of the study showed that a reduction in joint count is highly beneficial for performance. What that study left out was the perception of Level of Detail (LoD) of animations, and what the Just Noticeable Difference (JND) of a percentage decrease on the joint count would be. Thus motivating a study on peoples’ perception of animation quality. Objectives. The aim is to study the perception of LoD of animations in a game-like environment whilst doing a simple search-and-click task. Aiming to find the JND between animations with a different number of joints while the players are set with a task that does not involve interacting with the characters performing the animation. Methods. A psychophysical experiment was performed using a game implementation in Unity. Unity was chosen because it has easy access to develop scripts and a game environment. Furthermore, it has a marketplace where already made content can be downloaded and reused, which made the user study much easier to create. Results. A total of 85.71% did not see any difference between the different qualities used for the animations. The 14.29% that saw a difference in the animations all saw the difference between the lowest and the one next to the lowest quality animation. Conclusions. Three out of 21 participants were able to see a difference in the lowest animation quality, whilst no one saw any difference in the other qualities. Thus, people were not able to see a density reduction of up to 62.26% fewer joints for the chosen animation. Due to the low number of positive detections in the quality change of the animations, the JND could not be reliably computed.

<p>To render outdoor scenes with lots of vegetation in real time is a big challenge. This problem has important applications in the areas of visualization and simulation. Some progress has been made the last years, but a previously unsolved difficulty has been to combine high rendering quality with abundant variation in scenes. </p><p>I present a method to mathematically generate and render vegetation in real time, with implementation in the scene graph Gizmo3D. The most important quality of the method is its ability to render scenes with many unique specimens with very low aliasing. </p><p>To obtain real time performance, a hierarchical level-of-detail scheme (LOD- scheme) is used which facilitates generation of vegetation in the desired level- of-detail on the fly. The LOD-scheme is texture-based and uses textures that are common for all specimens of a whole species. The most important contribution is that I combine this LOD-scheme with the use of semi- transparency, which makes it possible to obtain low aliasing. </p><p>Scenes with semi-transparency require correct rendering order. I solve this problem by introducing a new method for approximate depth sorting. An additional contribution is a variant of axis-aligned billboards, designated blob, which is used in the LOD-scheme. Furthermore, building blocks consisting of small branches are used to increase generation performance.</p>

<p>Simulation of modern battlefield scenarios on consumer PC:s deal with a number of limitations, many of them related to the limited performance of a normal PC compared to workstations and servers. Specifically, the visualization of realistic large scale outdoor environments is problematic because of the large amount of data required to describe its contents. This becomes especially problematic in simulations of fast moving vehicles such as aircrafts, where one needs to maintain high frame rates while having high visual detail for orientation and targeting. This thesis proposes a method of generating realistic outdoor environments from actual geological data and then rendering it efficiently using an improved level of detail algorithm within a proprietary battle simulation framework. We also show how to integrate an open source Flight Dynamics Model (FDM) into the simulation framework for future hybrid simulations involving aircrafts.</p>

To render outdoor scenes with lots of vegetation in real time is a big challenge. This problem has important applications in the areas of visualization and simulation. Some progress has been made the last years, but a previously unsolved difficulty has been to combine high rendering quality with abundant variation in scenes. I present a method to mathematically generate and render vegetation in real time, with implementation in the scene graph Gizmo3D. The most important quality of the method is its ability to render scenes with many unique specimens with very low aliasing. To obtain real time performance, a hierarchical level-of-detail scheme (LOD- scheme) is used which facilitates generation of vegetation in the desired level- of-detail on the fly. The LOD-scheme is texture-based and uses textures that are common for all specimens of a whole species. The most important contribution is that I combine this LOD-scheme with the use of semi- transparency, which makes it possible to obtain low aliasing. Scenes with semi-transparency require correct rendering order. I solve this problem by introducing a new method for approximate depth sorting. An additional contribution is a variant of axis-aligned billboards, designated blob, which is used in the LOD-scheme. Furthermore, building blocks consisting of small branches are used to increase generation performance.

Rendering large terrains on large screens at interactive frame rates is a challenging area of computer graphics. In the last decade, real-time terrain rendering on large screens played a significant role in various simulations and virtual reality systems. To fulfill the demand of these systems, two software tools are developed. The first tool is a Terrain Editor that creates and manipulates large terrains. The second is a Multi-Display Viewer that displays the created terrains on multiple screens. Since the typical large terrains consist of many polygons, graphics boards might have difficulties in rendering the terrain at interactive frame rates. The common solution to this problem is to use terrain simplification without losing image quality. To this purpose, in this study, a paged level of detail mechanism that works with multiple threads is developed and integrated on multiple screen display systems to increase the performance of the high resolution systems.

We present a method for representing seams in clothing animation, and its application in simulation level of detail. Specifically we consider cloth represented as a regular grid of particles connected by spring-dampers, and a seam specified by a closed set of parametric trim curves in the cloth domain. Conventional cloth animation requires the tessellation of seams so that they are handled uniformly by the dynamics process. Our goal is a seam definition which does not constrain the attached clothing panels to be of the same resolution, or even constant resolution, while not being a hindrance to the dynamics process. We also apply our seams to cloth defined on a regular grid, as opposed to the irregular meshes commonly used with seams. The determination of particles interior to the cloth panel can be done using wellknown graphics operations such as scan-conversion. Due to the particle-based nature of the simulation, the dynamics approach combines easily with existing implicit and explicit methods. Finally, because the seams are resolution independent, the particle density per clothing panel can be adjusted as desired. This gives rise to a simple application of the given seams approach illustrating how it may be used for simulation level of detail.

<p>Denna rapport presenterar ett alternativt sätt att ljussätta terräng i datorgrafik. Tidigare modeller har vanligtvis byggt på lokal ljussättning, som inte tar hänsyn till kringliggande geometri, och har med en extra process approximerat effekten av ljusinteraktionen. Genom att använda sig av en teknik som kallas precomputed radiance transfer (PRT) kan man förberäkna hur en punkt interagerar med ljus för olika inkommande riktningar och undviker därmed att göra detta under programkörningen. Det är viktigt att denna teknik även fungerar tillsammans med level-of-detail (LOD) terrängrenderingsalgoritmer eftersom rendering av alla trianglar i terrängen för varje skärmuppdatering inte är optimalt för dagens grafikkort. Man vill därför representera den underliggande terrängen med fler trianglar närmare betraktaren och färre längre bort. Motiveringen till detta är att trianglar längre ifrån betraktaren kommer resultera i färre pixlar på skärmen, så att rendera större trianglar gör en liten visuell skillnad. Arbetet visar på att tekniken fungerar med LOD-terrängrenderingsalgoritmer med bra prestanda och visuell kvalitet.</p>

Syftet med denna studie var att visa vilka konsekvenser valet av en viss detaljnivå får vid simuleringen av ett system bestående av mjukvara, hårdvara och mekaniska komponenter. Till detta syfte ställdes två frågeställningar “Vad blir konsekvenserna av att lägga sig på en viss detaljnivå vid simulering av ett system bestående av mjukvara, hårdvara och mekaniska komponenter?” och “Hur påverkar detaljnivån när en simulering av ett system bestående av mjukvara, hårdvara och mekaniska komponenter ska verifieras och valideras?”. För att genomföra denna studie och uppnå syftet användes en vanligt förekommande metod för att utveckla simuleringsmodeller. Studien visade på att detaljnivån påverkar vilken information en modell kräver för att konstrueras och vilket användningsområde en modell har. Resultaten tyder på att informationen som finns tillgänglig om det system som en modell ska utvecklas för helt eller delvis begränsar vilken detaljnivå som modellen kan utvecklas på. Studien har begränsat undersökningen till en typ av växelspak och de två modeller som utvecklades ligger på en hög detaljnivå. Examensarbetet har även inte undersökt utvecklingen av modeller, enbart färdiga modeller. / The purpose of this study was to show what consequences a particular level of detail has on the simulation of a system consisting of software, hardware and mechanical components. In order to fulfil this purpose, the following questions were asked: “What are the consequences of using a certain level of detail in a simulation of a system containing software, hardware, and mechanical components?” and “How does the level of detail affect when a simulation of a system containing software, hardware, and mechanical components is verified and validated?”. To carry out this study, a common method for developing simulation models was used. The study showed that the level of detail will affect what information a model requires to be constructed and which uses a model has at that level of detail. The results suggest that the available information about the system that a model is developed for partially restricts the level of detail the model can have. The study is limited to a single type of shifter and all models have been developed at a high level of detail. The study also did not examine the development of models, and only studied finished models.