Projects

The difficulting of the projects below are ranked with between one and five stars, which indicate:

(*) - Implementation Problem. The project requires little original thought and the ideas have already been established in the literature. The project should be completely implemented, in stable release form, by the end of the semester.

(**) - Development Problem. The project will require some original thought by the student though some of the ideas have already been established by the literature. A working prototype should be completed by the end of the semester.

(***) - Advanced Development Problem. The project will require a lot of original thought by the student as the ideas are very recent and the literature is either very new are difficult to track down. A prototype that works at least for some cases should be provided by the end of the semester.

(****) - Research Problem. The project is a new idea that has not yet been published, and will depend on original thought by the student. Some results showing the problem can be solved should be shown by the end of the semester.

(*****) - Hard Research Problem. The project is a new idea and its solution will depend on brilliance on the part of the student. A careful study of the problem, with a proposed solution or a detailed explaination why the problem can't be solved should be given by the end of the semester.

• Random Numbers & Noise (*)
  Implement the random number functionality and noise functions for PGI, both at the geometry synthesis and shader level. Include the Worley Noise function. (**)
  Papers:  Lewis, Algorithms for Solid Noise Synthesis, Proc. SIGGRAPH 89; Worley, A Cellular Texture Basis Function, Proc. SIGGRAPH 96; Perlin, Improving Noise, Proc. SIGGRAPH 2002.
   
• Animation (*)
  Incorporate animation into the PGI specification.
  Paper:  Prusinkiewicz, Hammel and Mjolsness. Animation of Plant Development, Proc. of SIGGRAPH 93. 
   
• User-Interactivity (**)
  Add functionality to PGI to allow the user to select a PGI node by clicking an instance on the screen. Also allow the user to change the viewpoint interactively.
  
• View-Dependent Recursion Limits (***)
  If the bounding volumes project to an area smaller than the size of a pixel on the screen, render the bounding volume instead of the contents of the current PGI node. Might want to estimate the opacity of the PGI node (****). Might also want to estimate the BRDF of the PGI node (****).
  Papers: Neyret, A General and Multiscale Model for Volumetric Textures, Proc. Graphics Interface '95.
   
• Texture Synthesis on Surfaces (**)
  Add the ability to synthesize texture onto a surface. If it is procedurally-generated surface, then (****).
  Papers: Turk, Texture Synthesis on Surfaces, Proc. SIGGRAPH 2001 and Wei & Levoy, Texture Synthesis over Arbitrary Manifold Surfaces, Proc. SIGGRAPH 2001.
  
• Device Independent Procedural Shading (*****)
  Develop a shading language designed to target the latest generation of programmable shading hardware. The language should be flexible enough to express Renderman-like procedural shaders. The language should also be device independent, and not require the programmer to specifically target code for vertex shaders v. pixel shaders. In other words, the shader should be a single program and the compiler should determine where best to divide it between the vertex and pixel shader.
  
• Cg and the NVidia NV30 (***)
  The NV30 will be NVidia's next major graphics accelerator, and its support for procedural shading includes much expanded support for vertex and pixel shader programs. Develop a PGI rendering class that targets Cg and the NV30's programmable vertex and pixel shaders.
  Technology Overview: Cg and NV30 Emulation, NVidia.
   
• OpenGL 2.0 and the Wildcat VP (***)
  The OpenGL 2.0 proposal includes a procedural shading language designed for compilation into vertex and pixel shaders on modern graphics cards. Develop a PGI rendering class that targets OpenGL 2.0 and the Wildcat VP's programmable vertex and pixel shaders.
  Technology Overview: Wildcat VP product info and OpenGL 2.0Specification, 3DLabs
   
• RenderMonkey and the ATI Radeon 9700 (***)
  The ATI Radeon 9700 is a new GPU with extensive vertex and pixel shader support. RenderMonkey is a development environment designed to support cooperative work between a graphics programmer and an artist when developing procedural shaders for hardware implementation. Develop a PGI rendering class that targets the Radeon 9700's vertex and pixel shaders.
  Technology Overview: Programming the Radeon 9700and the RenderMonkey Toolsuite, ATI.
   
• Hardware Shadows (**)
  Incorporate hardware supported shadowing into the PGI rendering system.
  Technology Overview: NVidia Shadows Documents .
   
• Browser Plugin (**)
  Implement a renderer for PGI that allows it to work as a Browser plug-in. If you give your browser a URL ending in .pgi, then the result should be the output of the PGI program run on that URL. We should also be able to embed PGI files as IMG targets in HTML.
   
• Appearance Preserving Simplification (**)
  Automatically take the output of a PGI model and simplify its geometry while preserving its appearance as a texture map.
  Paper: Cohen, Olano and Manocha, Appearance Preserving Simplification, Proc. SIGGRAPH 98.
  Technology Overview: Normal Map Generator, NVSDK.
   
• Hardware bump mapping (**)
  Technology Overview: Bump Map Generation and Preview Tool, NVSDK.
   
• Hardware displacement mapping (***)
   
• Static bounding volumes (****)
  Automatically analyize a PGI node and construct a bounding volume that will work for all possible inputs. Assume the inputs are intervals. You can use interval arithmetic (**) or affine arithmetic (***), but a rotation by an interval amount creates a solid angle.
   
• Dynamic bounding volumes (*****)
  Automatically analyze a PGI node and construct a procedure that generates a bounding volume. The bounding volume should accept the same parameters as the PGI node, and construct a tight bounding volume for the specific instance of the PGI node.
   
• Static Analysis (*****)
  Apply static analysis to a PGI shading expression to predict its possible outcomes and construct a bounding volume for the PGI node without executing the node.
  Papers: Rice, Spatial Bounding of Self-Affine Iterated Function System Attractor Sets, Proc. Graphics Interface '96.
   
• Compile shading expressions into texture shaders (***)
  Construct a simple expression language like ESMTL and compile the expression into multitexture shaders.
   
• Implementing homomorphic BRDF (**)
  Given a BRDF (either as an expression or sampled) construct a multitexture shader using homomorphic factorization.
  Paper: McCool, Ang, & Ahmad, Homomorphic Factorizations of BRDFs for High-Performance Rendering , Proc. SIGGRAPH 2001.
  
• Automatic Computation of Reflectance Maps (***)
  Combine the BRDF and the environment map of an object into a reflectance map for PGI objects.
  Papers: W. Heidrich and H.-P. Seidel, Realistic, Hardware-accellerated Shading and Lighting, Proc. SIGGRAPH '99; and Cabral, Olano & Nemec, Reflection Space Image Based Rendering , Proc. SIGGRAPH 99.
   
• Meshing of procedural geometry (****)
  PGI constructs a blade of grass as a bunch of cylindrical segments. We would like to connect these segments into a single generalized cylinder mesh. How can PGI be modified to yield a mesh instead of a collection of disjoint geometric primitives?
   
• Context Sensitivity in the Scene Graph (*****)
  Explore how context sensitivity can be incorporated into a scene description language, such that nodes can behave differently based on properties of their parents, children and siblings.