Evaluating the transmittance between two points along a ray is a key component in solving the light transport through heterogeneous participating media and entails computing an intractable exponential of the integrated medium's extinction coefficient. While algorithms for estimating this transmittance exist, there is a lack of theoretical knowledge about their behaviour, which also prevent new theoretically sound algorithms from being developed. For this purpose, we introduce a new class of unbiased transmittance estimators based on random sampling or truncation of a Taylor expansion of the exponential function. In contrast to classical tracking algorithms, these estimators are non-analogous to the physical light transport process and directly sample the underlying extinction function without performing incremental advancement. We present several versions of the new class of estimators, based on either importance sampling or Russian roulette to provide finite unbiased estimators of the infinite Taylor series expansion. We also show that the well known ratio tracking algorithm can be seen as a special case of the new class of estimators. Lastly, we conduct performance evaluations on both the central processing unit (CPU) and the graphics processing unit (GPU), and the results demonstrate that the new algorithms outperform traditional algorithms for heterogeneous mediums.

Measured materials are rapidly becoming a core component in the photo-realistic image synthesis pipeline. The reason is that data-driven models can easily capture the underlying, fine details that represent the visual appearance of materials, which can be difficult or even impossible to model by hand. There are, however, a number of key challenges that need to be solved in order to enable efficient capture, representation and interaction with real materials. This paper presents two new data-driven BRDF models specifically designed for 1D separability. The proposed 3D and 2D BRDF representations can be factored into three or two 1D factors, respectively, while accurately representing the underlying BRDF data with only small approximation error. We evaluate the models using different parameterizations with different characteristics and show that both the BRDF data itself and the resulting renderings yield more accurate results in terms of both numerical errors and visual results compared to previous approaches. To demonstrate the benefit of the proposed factored models, we present a new Monte Carlo importance sampling scheme and give examples of how they can be used for efficient BRDF capture and intuitive editing of measured materials.

We present an overview and evaluation of a new, systematic approach for generation of highly realistic, annotated synthetic data for training of deep neural networks in computer vision tasks. The main contribution is a procedural world modeling approach enabling high variability coupled with physically accurate image synthesis, and is a departure from the hand-modeled virtual worlds and approximate image synthesis methods used in real-time applications. The benefits of our approach include flexible, physically accurate and scalable image synthesis, implicit wide coverage of classes and features, and complete data introspection for annotations, which all contribute to quality and cost efficiency. To evaluate our approach and the efficacy of the resulting data, we use semantic segmentation for autonomous vehicles and robotic navigation as the main application, and we train multiple deep learning architectures using synthetic data with and without fine tuning on organic (i.e. real-world) data. The evaluation shows that our approach improves the neural network’s performance and that even modest implementation efforts produce state-of-the-art results.

Data driven reflectance models using BRDF data measured from real materials, e.g. [Matusik et al. 2003], are becoming increasingly popular in product visualization, digital design and other applications driven by the need for predictable rendering and highly realistic results. Although recent analytic, parametric BRDFs provide good approximations for many materials, some effects are still not captured well [Löw et al. 2012]. Thus, it is hard to accurately model real materials using analytic models, even if the parameters are fitted to data. In practice, it is often desirable to apply small edits to the measured data for artistic purposes, or to model similar materials that are not available in measured form. A drawback of data driven models is that they are often difficult to edit and do not easily lend themselves well to artistic adjustments. Existing editing techniques for measured data [Schmidt et al. 2014], often use complex decompositions making them difficult to use in practice.

Traditional HDR capture has mostly relied on merging images captured with different exposure times. While this works well for static scenes, dynamic scenes poses difficult challenges as registration of differently exposed images often leads to ghosting and other artifacts. This chapter reviews methods which capture HDR-video frames within a single exposure time, using either multiple synchronised sensors, or by multiplexing of the sensor response spatially across the sensor. Most previous HDR reconstruction methods perform demoisaicing, noise reduction, resampling (registration), and HDR-fusion in separate steps. This chapter presents a framework for unified HDR-reconstruction, including all steps in the traditional imaging pipeline in a single adaptive filtering operation, and describes an image formation model and a sensor noise model applicable to both single-, and multi-sensor systems. The benefits of using raw data directly are demonstrated with examples using input data from multiple synchronized sensors, and single images with varying per-pixel gain.

With the development of modern image sensors enabling flexible image acquisition, single shot HDR imaging is becoming increasingly popular. In this work we capture single shot HDR images using an imaging sensor with spatially varying gain/ISO. In comparison to previous single shot HDR capture based on a single sensor, this allows all incoming photons to be used in the imaging, instead of wasting incoming light using spatially varying ND-filters, commonly used in previous works. The main technical contribution in this work is an  extension of previous HDR reconstruction approaches for single shot HDR imaging based on local polynomial approximations [15,10]. Using a sensor noise model, these works deploy a statistically informed filtering operation to reconstruct HDR pixel values. However, instead of using a fixed filter size, we introduce two novel algorithms for adaptive filter kernel selection. Unlike previous works, using  adaptive filter kernels [16], our algorithms are based on analysing the model fit and the expected statistical deviation of the estimate based on the sensor noise model. Using an iterative procedure we can then adapt the filter kernel according to the image structure and the statistical image noise. Experimental results show that the proposed filter de-noises the noisy image carefully while well preserving the important image features such as edges and corners, outperforming previous methods. To demonstrate the robustness of our approach, we have exploited input images from raw sensor data using a commercial off-the shelf camera. To further analyze our algorithm, we have also implemented a camera simulator to evaluate different gain pattern and noise properties of the sensor.

We present a new compressed sensing framework for reconstruction of incomplete and possibly noisy images and their higher dimensional variants, e.g. animations and light-fields. The algorithm relies on a learning-based basis representation. We train an ensemble of intrinsically two-dimensional (2D) dictionaries that operate locally on a set of 2D patches extracted from the input data. We show that one can convert the problem of 2D sparse signal recovery to an equivalent 1D form, enabling us to utilize a large family of sparse solvers. The proposed framework represents the input signals in a reduced union of subspaces model, while allowing sparsity in each subspace. Such a model leads to a much more sparse representation than widely used methods such as K-SVD. To evaluate our method, we apply it to three different scenarios where the signal dimensionality varies from 2D (images) to 3D (animations) and 4D (light-fields). We show that our method outperforms state-of-the-art algorithms in computer graphics and image processing literature.

Photo-realistic rendering of virtual objects into real scenes is one of the most important research problems in computer graphics. Methods for capture and rendering of mixed reality scenes are driven by a large number of applications, ranging from augmented reality to visual effects and product visualization. Recent developments in computer graphics, computer vision, and imaging technology have enabled a wide range of new mixed reality techniques including methods of advanced image based lighting, capturing spatially varying lighting conditions, and algorithms for seamlessly rendering virtual objects directly into photographs without explicit measurements of the scene lighting. This report gives an overview of the state-of-the-art in this field, and presents a categorization and comparison of current methods. Our in-depth survey provides a tool for understanding the advantages and disadvantages of each method, and gives an overview of which technique is best suited to a specific problem.

This thesis presents methods for photorealistic rendering of virtual objects so that they can be seamlessly composited into images of the real world. To generate predictable and consistent results, we study physically based methods, which simulate how light propagates in a mathematical model of the augmented scene. This computationally challenging problem demands both efficient and accurate simulation of the light transport in the scene, as well as detailed modeling of the geometries, illumination conditions, and material properties. In this thesis, we discuss and formulate the challenges inherent in these steps and present several methods to make the process more efficient.

In particular, the material contained in this thesis addresses four closely related areas: HDR imaging, IBL, reflectance modeling, and efficient rendering. The thesis presents a new, statistically motivated algorithm for HDR reconstruction from raw camera data combining demosaicing, denoising, and HDR fusion in a single processing operation. The thesis also presents practical and robust methods for rendering with spatially and temporally varying illumination conditions captured using omnidirectional HDR video. Furthermore, two new parametric BRDF models are proposed for surfaces exhibiting wide angle gloss. Finally, the thesis also presents a physically based light transport algorithm based on Markov Chain Monte Carlo methods that allows approximations to be used in place of exact quantities, while still converging to the exact result. As illustrated in the thesis, the proposed algorithm enables efficient rendering of scenes with glossy transfer and heterogenous participating media.

En av de största utmaningarna inom datorgrafik är att syntetisera, eller rendera, fotorealistiska bilder. Fotorealistisk rendering används idag inom många tillämpningsområden såsom specialeffekter i film, datorspel, produktvisualisering och virtuell verklighet. I många praktiska tillämpningar av fotorealistisk rendering är det viktigt att kunna placera in virtuella objekt i fotografier, så att de virtuella objekten ser verkliga ut. IKEA-katalogen, till exempel, produceras i många olika versioner för att passa olika länder och regioner. Grunden till de flesta bilderna i katalogen är oftast densamma, men symboler och standardmått på möbler varierar ofta för olika versioner av katalogen. Istället för att fotografera varje version separat kan man använda ett grundfotografi och lägga in olika virtuella objekt såsom möbler i fotot. Genom att på det här sättet möblera ett rum virtuellt, istället för på riktigt, kan man också snabbt testa olika möbleringar och därmed göra ekonomiska besparingar.

Den här avhandlingen bidrar med metoder och algoritmer för att rendera fotorealistiska bilder av virtuella objekt som kan blandas med verkliga fotografier. För att rendera sådana bilder används fysikaliskt baserade simuleringar av hur ljus interagerar med virtuella och verkliga objekt i motivet. För fotorealistiska resultat kräver simuleringarna noggrann modellering av objektens geometri, belysning och materialegenskaper, såsom färg, textur och reflektans.

För att de virtuella objekten ska se verkliga ut är det viktigt att belysa dem med samma ljus som de skulle ha haft om de var en del av den verkliga miljön. Därför är det viktigt att noggrant mäta och modellera ljusförhållanden på de platser i scenen där de virtuella objekten ska placeras. För detta använder vi High Dynamic Range-fotografi, eller HDR. Med hjälp av HDR-fotografi kan vi noggrant mäta hela omfånget av det infallande ljuset i en punkt, från mörka skuggor till direkta ljuskällor. Detta är inte möjligt med traditionella digitalkameror, då det dynamiska omfånget hos vanliga kamerasensorer är begränsat. Avhandlingen beskriver nya metoder för att rekonstruera HDR-bilder som ger mindre brus och artefakter än tidigare metoder. Vi presenterar också metoder för att rendera virtuella objekt som rör sig mellan regioner med olika belysning, eller där belysningen varierar i tiden. Metoder för att representera spatiellt varierande belysning på ett kompakt sätt presenteras också. För att noggrant beskriva hur glansiga ytor sprider eller reflekterar ljus, beskrivs också två nya parametriska modeller som är mer verklighetstrogna än tidigare reflektionsmodeller. I avhandlingen presenteras också en ny metod för effektiv rendering av motiv som är mycket beräkningskrävande, till exempel scener med uppmätta belysningsförhållanden, komplicerade  material, och volumetriska modeller som rök, moln, textiler, biologisk vävnad och vätskor. Metoden bygger på en typ av så kallade Markov Chain Monte Carlo metoder för att simulera ljustransporten i scenen, och är inspirerad av nyligen presenterade resultat inom matematisk statistik.

Metoderna som beskrivs i avhandlingen presenteras i kontexten av fotorealistisk rendering av virtuella objekt i riktiga miljöer, då majoriteten av forskningen utförts inom detta område. Flera av de metoder som presenteras i denna avhandling är dock tillämpbara inom andra domäner, såsom fysiksimulering, datorseende och vetenskaplig visualisering.

Accurate and efficient simulation of light transport in heterogeneous participating media, such as smoke, clouds and fire, plays a key role in the synthesis of visually interesting renderings for e.g. visual effects, computer games and product visualization. However, rendering of scenes with heterogenous participating with Metropolis light transport (MLT) algorithms have previously been limited to primary sample space methods or using biased approximations of the transmittance in the scene. This paper presents a new sampling strategy for Markov chain Monte Carlo (MCMC) methods, e.g. MLT, based on pseudo-marginal MCMC. Specifically, we show that any positive and unbiased estimator of the target distribution can replace the exact quantity to simulate a Markov Chain with a stationary distribution that has a marginal which is the exact target distribution of interest. This enables us to evaluate the transmittance function with recent unbiased estimators which leads to significantly shorter rendering times. Compared to previous work, relying on (biased) ray-marching for evaluating transmittance, our method enables simulation of longer Markov chains, a better exploration of the path space, and consequently less image noise, for a given computational budget. To demonstrate the usefulness of our pseudo-marginal approach, we compare it to representative methods for efficient rendering of anisotropic heterogeneous participating media and glossy transfer. We show that it performs significantly better in terms of image noise and rendering times compared to previous techniques. Our method is robust, and can easily be implemented in a modern renderer.