This presentation describes construction of a large city environment for DreamWorks Animation's "Kung Fu Panda 2", including how the chosen technique was driven at every step by the film's design language and story needs to deliver an epic setting within reasonable time constraints.

“Legend of the Guardians: the Owls of Ga'Hoole” evokes an epic nocturnal owl world, eschewing a cartoon aesthetic in favour of a dark, operatic style. The emotional drama of the story was more important than maintaining a sense of realism or naturalism.

Blue Sky Studios produced a large number of art-directed cloud environments for "Rio", using optimized creation, management, compositing, and ray-tracing technologies and techniques.

This talk illustrates improvements and additions to the render pipeline and the tools pipeline to enable fast, slick game play mixed with cinematic explosions and dynamic events to produce the most visceral, exciting racing title possible.

A method for integrating disparate traffic-simulation techniques that allows for interactive, large-scale visualizations of traffic flow without compromising quality. The approach uses both agent-based and continuum techniques to balance user needs and achieve a high level of performance.

The ability to direct a viewer’s gaze about a scene has important applications in computer graphics and data visualization. This talk describes an experiment designed to study the impact of subtle gaze manipulation on short-term spatial information recall.

A technique for high-resolution correspondence of independent facial expression scans, incorporating inverse rendering, GPU computation, and real-time user direction during the optimization; corresponded expression data are mapped into a common domain for use in realistic facial animation.

A discussion of how the "Cars 2" character and crowd departments co-developed a strategy to maximize show asset efficiency and agent behavioral flexibility into a production pipeline doubling the shot throughput.

A new modular architecture for the film "Rio" enabled artists to procedurally generate and animate vast quantities and varieties of detail, from flocks of birds in windswept jungles to thousands of costumed characters at Carnaval, while improvements to voxel rendering and sparse octree geometry caches facilitated ray-tracing this complexity.

Blue Sky Studios developed an efficient workflow for staging the carnival scenes in the "Rio" and created a way of rendering them in a ray tracer such that a scenes with more than 30,000 characters fit in the memory of a single computer.

This novel method to physically simulate flying motions that imitate real bird flight generates physically plausible and apparently realistic bird behaviors based on real wingbeat data.

This talk discusses building information modeling (BIM) tools as an opportunity for design generation, validation, and implementation. It specifically focuses on parametric modeling and constraint-based design explorations and construction assemblies.

This talk focuses on strategies for generative-design validation with the use of digital simulations, particularly dynamics-based modeling tools. It focuses on tools that employ rigid-body or soft-body dynamics, such as cloth simulations, forward and inverse kinematics (FK/IK), and particle interactions.

This work flow for modeling and animating cobwebs on antique sets includes a tool set for artists to model cobwebs, a physics-based solver, and a procedural-modeling engine.

An in-depth look at the techniques used to light and render Tokyo in "Cars 2".

Lighting Metro City at night required solving the lighting challenge known as the "many lights" problem. This sketch describes how the lighting department solved this challenge.

A technical and production-oriented talk describing the tile-based lighting architecture and deferred rendering optimizations developed to achieve a visual target with hundreds of real-time lights per scene for Need for Speed: The Run at 30 frames per second on PlayStation3.

A new, high-quality, and efficient single-image upscaling method that exploits local self-similarities in the data and reduces search time. The method is implemented using novel non-dyadic filter banks that allow real-time processing when converting low-resolution video formats into HD standards.

A scalable and temporal coherent video retargeting method that factors the problem into spatial and motion components. The sub-problem sizes are proportional to a frame resolution, and they can be solved in parallel. The system dramatically reduces computational time and memory consumption without compromising quality.

A robust and efficient approach to video stabilization that achieves high-quality camera motion for a wide range of videos and is sufficiently practical for consumer applications.

This paper presents a method for reducing undesirable tonal fluctuations in video: minute changes in tonal characteristics, such as exposure, color temperature, brightness, and contrast in a sequence of frames. These distracting fluctuations are typically caused by the camera's automatic setting adjustment.

A metric for predicting visible differences (discrimination) and image quality (MOS) in high-dynamic-range images. The metric is based on new contrast sensitivity measurements and calibrated against several datasets. The visibility predictions are shown to be improved compared to the original HDR-VDP and VDP metrics.

An optical architecture for HDR imaging that allows simultaneous capture of high-, medium-, and low-exposure images, and an HDR merging algorithm that avoids undesired artifacts. This paper explains implementation of a prototype HDR-video system and present sresults from the acquired HDR video.

A perceptually based algorithm for modeling the color shift that occurs for human viewers in low-light scenes. Known as the Purkinje effect, this color shift occurs as the eye transitions from photopic, cone-mediated vision in well-lit scenes to scotopic, rod-mediated vision in dark scenes.

Changing the color of an object is a basic image-editing operation, but a high-quality result must also preserve natural shading. This approach decomposes illumination into direct lighting and indirect illumination from individual materials to consistently modify material colors and their associated interreflections.

This paper presents:

• A novel spatio-temporal anti-aliasing algorithm with essentially noise-free results. Visibility is solved for along line samples in screen space and in time.
• A new method for ambient occlusion based on line samples/ The method can also handle motion-blurred ambient occlusion.

A new frequency analysis and sheared filter for computing shadows cast by complex occluders. The approach uses ray tracing to sparsely sample the scene and a sheared 4D filter to share shadow samples between neighboring pixels.

In this paper, a method for reconstructing high-quality images from sparse stochastic samples is applied to simultaneous motion blur, depth of field, and soft shadows.

This paper introduces a homogeneous transform that reduces the computation required to determine the set of points or primitives inside a tetrahedral volume. It describes how application of this transform can improve the efficiency of soft shadows and defocus blur computations.

A new method for generating spatially varying stochastic-point distributions with blue-noise spectrum in linear time. The method relies on a novel multi-scale strategy that draws samples from an interacting particles statistical model that formalizes the problem.

This paper solves the problem of generating a uniform Poisson-disk sampling that is both maximal and unbiased over bounded non-convex domains. The method is fast, requires little memory, and is guaranteed to terminate without any bias. The paper also demonstrates a parallel GPU implementation that performs MPS in a bias-free fashion.

Existing spatial and spectral methods for analyzing sample distributions are restricted to uniform domains. This paper presents differential domain analysis for general non-uniform scenarios including adaptive, anisotropic, and surface sampling, by reformulating standard Fourier analysis that depends on sample locations into an equivalent form that depends on only location differentials.

Existing noise functions either introduce discontinuities of the solid noise at sharp edges (wavelet noise, Gabor noise) or result in detail loss when anti-aliased (Perlin noise, wavelet noise). This paper presents a new noise function that preserves continuity over sharp edges and supports high-quality anti-aliasing.

A grammar-based framework for converting 3D furniture models to separate parts and connectors that can be physically built, which allows nonprofessional users to create products that they can use themselves by "printing" 3D furniture models.

This paper introduces a framework for automatic generation of furniture arrangement, considering human factors such as accessibility, visibility, pathways, etc. The framework generates arrangements for a variety of scenarios.

An interactive furniture layout system that assists users by suggesting furniture arrangements based on interior design guidelines. Results demonstrate that the suggestion-generation functionality measurably increases the quality of furniture arrangements produced by participants with no prior training in interior design.

Procedural extrusion is a new, easy-to-use tool that combines arbitrary plans and profiles to create a wide range of architectural forms. This paper demonstrates the technique's application to procedural modeling of cityscapes.

This paper brings artistic control to grammar-based procedural models via Markov-chain Monte Carlo inference. Given a grammar and a high-level specification, the algorithm computes a production from the grammar that conforms to the specification. The method is demonstrated on models of trees, cities, buildings, and Mondrian paintings.

This simple framework for progressive processing of high-resolution images with minimal resources implements an adaptive, multi-resolution solver for gradient-based image processing that is capable of real-time seamless cloning for images that are hundreds of gigapixels in size.

A new, unified technique to perform general edge-sensitive editing operations on n-dimensional images and videos efficiently. Introduction of a generalized geodesic distance transform provides the unifying framework to efficiently address several editing operations such as de-noising, segmentation, and non-photorealistic rendering.

This image-based model for matting and compositing transparent and refractive objects from single images encodes plausible refractive properties of a transparent object and its observed specularities and transmissive properties. The method simulates compound compositing, Fresnel effect, scene depth, and caustic shadows from the extracted model.

A nonlinear revision-control system for images that supports review, diff, addition, branching, merging, and conflict resolving, and facilitates artistic creation in common image editing and digital sketching tasks. The core idea is a DAG structure that represents the spatial, temporal, and semantic relationships between operations.

A general framework for performing curvature-aware geometry processing through solution of a screened Poisson equation and a parallel streaming implementation for constructing and solving the associated linear system at interactive rates.

The double insertion, non-uniform, stationary subdivision surface (DINUS) generalizes both the non-uniform, bicubic spline surface and the Catmull-Clark subdivision surface. DINUS allows arbitrary knot intervals on the edges, allows incorporation of special features, and provides limit point as well as limit normal rules.

By interpolating basis functions built on the Sinc function with a Guassian multiplier, this technique constructs freeform interpolatory curves and surfaces through scattered data points.