Siggraph 2007 Papers
This is my short list of most-interesting papers from Siggraph 2007. These papers either advance the art in a field that I am particularly interested in such as character animation, procedural modeling, or high quality lighting, or simply catch my fancy. The numbers refer to the chapter index in the paper proceedings. In all cases, I have tried to link the project website; if there isn't one, then the principal researcher's website; if not that, then a pdf of the paper. 2. Solid Texture Synthesis from 2D Exemplars. Johannes Kopf, Chi-WIng Fu, Daniel Cohen-Or, Oliver Deussen, Dani Lischinksi, Tien-Tsin Wong. Re-weighted least squares are used to minimize the differences between an exemplar texture and an arbitrary slice through a volume. Compared to previous approaches this new approach has several advantages including the ability to retain correlations across multiple dimensions (for example, color channels remain correlated). 7. Near-Optimal Character Animation with Continuous Control, Adrien Treuille, Yongjoon Lee, Zoran Popovic. Low dimensional basis controllers are derived to allow responsive transitions between animation clips spanning a very large move tree. Parameterization of the controller utilizes a key insight that foot plants can be used to synchronize transitions between clips. 15. Continuous Collision Detection for Articulated Models using Taylor Models and Temporal Culling, Xinyu Zhang, Stephane Redon, Minkyong Lee, Young J. Kim. Taylor models allow tight fitting bounds over articulated segments of a model. These tight fitting bounds avoid the exponential growth of bounds under interval arithmetic, and as a result, a culling phase discards most potential collisions, allowing a huge speed up. 18. Line Drawings via Abstracted Shading, Yunjin Lee, Lee Markosian, Seungyong Lee, John F. Hughes. Rendering a tone map and then doing line detection via gradient search on the GPU creates a very clean toon shader. 23. Algebraic Point Set Surfaces. Gael Guennebaud, Markus Gross. Oriented points in a point set yield much better reconstructed surfaces than unoriented points. 25. The Lightspeed Automatic Interactive Lighting Preview System. Jonathan Ragan-Kelley, Charlie Kilpatrick, Brian W. Smith, Doug Epps, Paul Green, Christophe Hery, Fredo Durand. A Renderman shader is split into a constants and dynamics phase. The constants phase is baked into deep framebuffers, and the dynamic part of the computation is rewritten as Cg allowing for interactive close to final rendering once the precomputation phase is complete. 26. Matrix Row-Column Sampling for the Many-Light Problem. Milos Hasan, Fabio Pellacini, Kavita Bala. Radiative transfer is mapped as the rows and columns of a huge matrix uploaded to the GPU allowing renders of scenes with hundreds of thousands of lights in only a few seconds. 29. Image-Guded Maze Construction. Jie Xu, Craig S. Kaplan. Given an image, a maze of desired complexity is generated. 41. FiberMesh: Designing Freeform Surfaces with 3D Curves. Andrew Nealen, Takeo Igarashi, Olga Sorkine, Marc Alexa. The further evolution of Teddy-like systems. The key difference here being that user sketched lines are retained, and the surface generated as a constrained optimization. 45. Plushie: An Interactive Design System for Plush Toys. Yuki Mori, Takeo Igarashi. Teddy like system is expanded with a physical simulation allowing users to generate stitching patterns for toys which can be sewn together into plushies IRL. 46. Curl-Noise for Procedural Fluid Flow. Robert Bridson, Jim Hourihan, Marcus Nordenstam. An exactly compressible turbulent velocity noise field is derived from Perlin noise, based on the vector calculus identity that the curl of a smooth potential is automatically divergence free. This allows the construction of particle flow fields that look just like fluid flow. Code available at website. 59. Eikonal Rendering: Efficient Light Transport in Refractive Objects. Ivo Ihrke, Gernot Ziegler, Art Tevs, Christian Theobalt, Marcus Magnor, Hans-Peter Siedel. An intriguing GPU based render method where patch particles are dispatched through a volume texture as an expanding wavefront, leaving energy where they interact with material in the volume. I expect wavefront based rendering to be a significant new research direction, having lain mostly dormant since the early 1980's. 61. Implicit Visibility and Antiradiance in Interactive Global Illumination. Carsten Dachsbacher, Marc Stamminger, George Drettakis, Fredo Durand. Light is transported implicitly without considering occlusion. Lights that are occluded contribute an antiradiance term from the occluding surface. This insight allows a large acceleration of the light transport calculation hinting at the beginning of real time GI. 63. Symmetrization. Niloy J. Mitra, Leonidas J. Guibas, Mark Pauly. A set of closed-form solutions for optimal symmetry calculations yields an algorithm for symmetrization of a mesh. This has applications for correspondence computations as well. 65. Geometry of Multi-Layer Freeform Structures for Architecture. Helmut Pottmann, Yang Liu, Johannes Wallner, Alexander Bobenko, Wenping Wang. Optimization of freeform structures given optimal beam length and node geometry yields new architectural structures. 68. Multi-Aperature Photography. Paul Green, Wenyang Sun, Wojciech Matusik, Fredo Durand. Multiple CCDs capture light from several aperature sizes simultaneously via a beam splitter in an attempt to create an all-focus camera. The main drawback of this system is a rather elaborate fixed apparatus, the use of which is difficult to imagine outside of a laboratory setting. 69. Dappled Photography: Mask Enhanced Cameras for Heterodyned Light Fields and Coded Aperature Refocusing. Ashok Veeraraghavan, Ramersh Raskar, Amit Agrawal, Ankit Mohan, Jack Tumblin. This version of the all-focus camera uses a coded mask to capture light fields which are recovered in post processing from the Fourier domain. 70. Image and Depth from a Conventional Camera with a Coded Aperature. Anat levin, Rob Fergus, Fredo Durand, William T. Freeman. An aperature mask is placed in the lens of a camera and blurs are discriminated versus the mask to create an all-focus camera. This system seems the most practical of those presented, but suffers from light loss due to the fact that the aperature mask lets much less light through than a conventional aperature. 71. Harmonic Coordinates for Character Articulation. Pushkar Joshi, Mark Meyer, Tony DeRose, Brian Green, Tom Sanocki. Mesh vertices are bound locally to best locations in a containing deformation lattice yielding highly posable characters without a traditional joint structure. 72. Automatic Rigging and Animation of 3D Characters. Ilya Baran, Jovan Popovic. Standardized character rigs are embedded in arbitrary meshes after automatic analysis. Released as a package named Pinocchio. Code available at website. 73. Real-Time Enveloping with Rotational Regression. Robert Y. Wang, Kari Pulli, Jovan Popovic. Deformation gradients are used in preference to linear blend skinning resulting in realtime character deformations that do not exhibit the typical collapse artifacts of LBS. This method relies on precomputed mesh poses that can be optimized against to create the gradient domain mapping. 78. Sampling with Polyominoes. Victor Ostromoukhov. Polyomino tilings are shown to have blue-noise characteristics when sample points are picked at the centers of tiles. 79. Stochastic Simplification of Aggregate Detail. Robert L. Cook, John Halstead, Maxwell Planck, David Ryu. Given an object such as a bush, or the hair on a rat, features are stochastically dropped from the representation according to projected screen extent. Features are also emphasized to retain an impression of the original aggregate detail. This allows objects that would normally not fit into memory (such as the hairs of thousands of rats) to be rendered. 85. Image-based Procedural Modeling of Facades. Pascal Muller, Gang Zeng, Peter Wonka, Luc Van Gool. Given a photograph of a building facade, a shape grammar is automatically derived. 87. Image-based Tree Modeling. Ping Tan, Gang Zeng, Jingdong Wang, Sing Bing Kang, Long Quan. In this procedural tree modeling approach, images are segmented, visible and occluded branches are reconstructed, and then leaves are introduced guided by an extracted 3D point cloud approximating the leaf volume. 88. Approximate Image-Based Tree-Modeling using Particle Flows. Boris Neubert, Thomas Franken, Oliver Deussen. Image slices of trees are used to approximate spatial densities which guide particles to grow in the shape of branches that fill the density field. 89. Fast Triangle Reordering for Vertex Locality and Reduced Overdraw. Pedro V. Sander, Diego Nehab, Joshua Barczak. A new triangle ordering algorithm is introduced which improves post-transform vertex cache coherency. 97. Bubbling and Frothing Liquids. Paul W. Cleary, Soon Hyoung Pyo, Mahesh Prakash, Bon Ki Koo. Heterogeneous SPH introduces bubbles to a more traditional SPH simulation; the bubbles can aggregate and grow as they rise forming a foamy head. 99. Wave Particles. Cem Yuksel, Donald H. House, John Keyser. Wave surface particles are introduced which allow for rapid and realistic simulation of wave surface interactions. 105. SIMBICON: Simple Biped Locomotion Control. KangKang Yin, Kevin Loken, Michiel van de Panne. Feedback Error Learning is applid to manual controller design to evolve a lightweight set of dynamic controllers that can reproduce various biped locomotion strategies, and emulate exemplars obtained from motion capture. Once the controller has been constructed it can be parameterized for interactive control.