With these new techniques for rendering translucent materials such as human skin, modified diffusion theory and a new quantized-diffusion method derive efficient and accurate scattering functions for both offline and real-time rendering.

A novel, comprehensive theory of volumetric radiance estimation that leads to new insights and includes all previous radiance estimates as special cases. The authors use this theory to develop a more efficient rendering method for participating media based on "photon beams".

A novel formulation of progressive photon mapping based on a probabilistic derivation. Unlike the original progressive photon-mapping algorithm, this method does not require maintenance of local photon statistics. It extends progressive photon mapping to arbitrary kernels and includes stochastic ray tracing and volumetric photon mapping.

A novel heuristic ray-reordering method for incoherent secondary rays. Its benefits are demonstrated in two stochastic rendering methods and different cache levels.

This paper proposes a new technique for exploring unorganized collections of 3D shapes. The approach represents the variability in the collection as a constrained deformation model of a base template shape, which the user can directly manipulate to explore the set of shapes.

This paper shows how to represent scenes as graphs that encode models and their semantic relationships. It then defines a kernel between these relationship graphs that compares common virtual substructures in two graphs and captures the similarity between their corresponding scenes.

A probabilistic model of shape structure that encodes semantic and geometric relationships among shape components. The model is trained on a shape database and is used to present relevant components during interactive 3D modeling.

Using heat diffusion kernels as "geometric words", this technique employs the "bag of features" paradigm to construct invariant shape descriptors. The descriptors provide state-of-the-art results on large-scale shape retrieval benchmarks.

An exact and efficient contributing-vertices-based algorithm for the Minkowski sum computation of a non-convex-convex pair of polyhedra. The algorithm handles non-manifold situations and extracts polyhedral holes of the sum. Several results

MeshFlow is an interactive system for visualizing polygonal mesh construction sequences. It hierarchically clusters editing operations visualized with graphical annotations to provide levels of detail on demand. MeshFlow was tested on five mesh sequences and evaluated with a case study using modeling students.

LR is a compact data structure for representing the connectivity of manifold triangle meshes. It supports constant-time incidence and adjacency queries while storing only one integer reference per triangle. Its greedy, linear-time construction involves ordering the vertices along a nearly Hamiltonian cycle.

Previously, only the normal component of the surface was known. This method of deriving a formula for the velocity of a time-evolving implicit surface applies the normal component to the problem of tracking particles on the surface and the problem of motion-blurring implicit surfaces.

This computational stereo camera system closes the control loop to handle traditionally difficult shots. It is completely programmable, so complex shots can easily be scripted and adjusted depending on the scene and artistic preferences.

A camera-projector system casts distinct grid patterns onto a scene to generate a spatial distribution of point reflections. These point reflections relay a relative measure of defocus that is utilized in post-processing to generate a highlighted depth-of-field photograph.

This method optimizes auto-multiscopic displays composed of compact volumes of light-attenuating material. Inexpensively fabricated by stacking transparencies, the attenuators recreate a light field when illuminated by a backlight. Tomographic optimization resolves inconsistencies between views, leading to brighter, higher-resolution 3D displays with extended depth of field and improved dynamic range.

This paper summarizes a psychophysical experiment to determine human sensitivity to binocular disparity on modern 3D stereo equipment. The obtained data were used to formulate a model of perceived disparity that lays the foundations for effective disparity compression, retargeting, personalized stereo, hybrid stereo images, and a perceived disparity-difference metric.

A novel method for increasing the efficiency of stochastic rasterization of motion and defocus blur. Contrary to earlier approaches, the method is efficient even with the low sampling densities commonly encountered in real-time rendering, while allowing the use of arbitrary sampling patterns for optimum image quality.

A generalized approach to decoupling shading from both visibility and geometry sampling in graphics pipelines. The technique enables stochastic supersampling of motion and defocus blur and other effects at reduced shading cost, without requiring micropolygon rendering. It can be thought of as a generalization of GPU-style multi-sample anti-aliasing (MSAA).

Spark improves on current real-time shading languages by supporting better software development practices. It allows the logical features of a real-time shader to be expressed as independent, composable modules, even where a feature cross-cuts multiple stages of the rendering pipeline.

An efficient method to render lens flare of complex optical systems. Based on a combination of traditional ray tracing and rasterization, our technique captures the intricate dynamics of real flare phenomena, while spanning a continuum between accuracy and performance, physical realism and artistic expression.

Deformation by linear blending dominates practical use as the fastest approach for transforming 2D and 3D shapes. This paper develops "bounded biharmonic weights" that produce smooth and intuitive deformations for any combination of points, bones, and cages of arbitrary topology, making control of deformations simple and flexible.

This method for inter-surface mapping generates a map between two surfaces by blending multiple low-dimensional (conformal) maps with smoothly varying weights. Performance is evaluated on a diverse dataset and compared our resulting maps to several state-of-the-art methods.

A new distance measure on manifolds based on the biharmonic differential operator that provides a nice trade-off between nearly geodesic distances for small distances and global shape-awareness for large distances.

An algorithm for photo-inspired creative 3D modeling where the user draws inspiration from a single photograph and the modeling process is supported by an available set of candidate objects.

This method for restoring antialiased edges that are damaged by certain types of nonlinear image processing operations selectively adjusts the local gradients in affected regions of the filtered image so that they are consistent with those in the original image.

This paper demonstrates state-of-the-art edge-aware image processing using Laplacian pyramids. The advantage of this approach is its conceptual simplicity and flexibility, based on simple point-wise nonlinearities; no optimization or post-processing is required. The method produces consistently high-quality results, without degrading edges or introducing halos.

A new approach for performing high-quality edge-preserving filtering of images and videos in real time.

A new efficient method for joint recovery of both reliable local dense correspondences and a global parametric color transformation model between two images. The paper demonstrates several applications of this method for automatic example-based enhancement of photographs that share some content.

A method to align locally detected RANSAC primitives using global relations, which are learned and conformed to using a combination of graph reduction and constrained optimization. The paper demonstrates applications on synthetic and scanned data, even under structured noise and anisotropic sampling.

The articulated global registration algorithm presented in this work aligns multiple range scans simultaneously to reconstruct a full poseable 3D model. Solving for surface motion using linear-blend skinning and automatically modeled joints allows users to interactively manipulate the resulting 3D model and create new animations.

In this lobe-based tree representation for modeling trees, the tree’s foliage is abstracted into compact canonical geometry structures, called lobe-textures. The method introduces techniques to encode a given tree as lobe-based representation and decode the representation and synthesize a fully detailed tree model.

With this ℓ1-sparse reconstruction of piecewise smooth objects, common objects can be characterized by a small number of features that introduce sparsity into the model. The reconstruction gives rise to shapes that consist mainly of smooth modes, with the objective function residual concentrated near sharp features.

A new real-time Eulerian fluid simulation that uses a hybrid grid of regular cubic cells on top of a layer of tall cells above an arbitrary terrain. The method includes a novel multigrid method and several fluid solver modifications, and achieves 30fps in real-world scenarios on a modern GPU.

A two-scale method for particle-based fluids that allocates computing resources to regions of the fluid where complex flow behavior emerges. This method produces similar surface details as the single-resolution simulation while improving the efficiency.

This paper presents a method for achieving non-reflecting boundaries when animating free surface fluids. The boundaries are based on the perfectly matched layer (PML) approach and are intended to allow localized high-quality simulations of large fluid domains such as oceans.

To efficiently obtain natural-looking liquid simulations subject to art direction, this method processes approved low-resolution geometry into a "guide shape" just below the liquid surface. The final high-resolution simulation runs just a surface layer constrained by the guide, with benefits for both speed and control.

A technique for synthesizing synchronized sounds for physically based fire animations. A hybrid method simulates low-frequency flame sounds using a physics-based approach, and then uses bandwidth extension and sound-texture synthesis to synthesize fiery high-frequency details.

This new approach for acquiring high-fidelity 3D facial performances with realistic dynamic wrinkles and fine facial details leverages state-of-the-art motion capture technology and advanced 3D scanning technology of facial-performance acquisition.

A new technique for high-quality facial performance capture that leverages the repetitive structure of face motions to automatically locate frames with similar expressions, called anchor frames. High-resolution geometry is reconstructed and temporal motion is propagated in parallel using robust image-space matching between the anchor frames.

A linear face modelling approach that generalizes better to unseen data and allows click-and-drag interaction for animation. The model consists of a collection of linear submodels independently trained but related through shared boundaries. It is demonstrated on dense-face motion capture data.

This paper presents a system for performance-based character animation that enables any user to control the facial expressions of a digital avatar in real-time using a non-intrusive, commercially available 3D sensor.

ShadowDraw, a system for guiding the freeform drawing of objects. As the user draws, ShadowDraw dynamically updates a shadow image underlying the user's strokes. The shadows suggest object contours to guide the drawing process.

OverCoat is a 3D painting system that allows the user to treat the full 3D space as a canvas. Paint strokes applied in 2D are embedded in a 3D scalar field that is shaped by proxy geometry, and rendered such that the result always looks like a 2D digital painting.

Designing and shading volumetric effects in feature-films is a difficult problem. This system, modeled after the way expert artists think about participating media, allows for intuitive control and authoring of both physically accurate and non-physical, artistic volumetric effects.

This noise function suitable for non-photorealistic rendering creates random variation that moves with the objects in a scene.

A new method for simulating realistic behavior of highly deformed, interacting solids without requiring explicit representation of the objects themselves.

A new and efficient algorithm for skinning skeleton-driven physically based elasticity. The algorithm and implementation are designed for efficiency with parallelism, matrix-free solvers, and branch-free vectorized and multi-threaded SVD computation. The paper demonstrates the system on real-world examples with contact and collisions.

This asynchronous mode-level-adaptive integrator with modified staggered projections synthesizes realistic multibody contact sounds. It resolves modal vibrations in both collision and frictional contact processing stages, thereby enabling non-rigid sound phenomena such as micro-collisions, vibrational energy exchange, and contact damping.

This paper focuses on the challenging problem of simulating frictional contact between thin elastic rods. Following the seminal work by Alart and Curnier in contact mechanics, this work simultaneously models contact and exact Coulomb friction as a root-finding problem that can be simply solved using a nonsmooth Newton algorithm.

A new framework for the simulation of constrained strands, which combines Lagrangian and Eulerian approaches. The two contributions are the reduced node, whose degrees of freedom precisely match the constraint, and the Eulerian node, which allows constraint handling that is independent of the initial discretization of the strand.