Sunteți pe pagina 1din 3

Written Assignment #5 Math for Computer Graphics

CSE 458 Computer Graphics, Fall 2010 Release Date: Tuesday, November 29, 2011 Due Date: Wednesday, December 7, 2011 The work must be your own. Write up the answers in your own words, explain all of the steps of reasoning and be neat! Also, please cite any external references you use (other than the textbook) to come up with your answers. 1. Interpolation List at least 3 dierent uses of an interpolation method in computer graphics. Explain how it works in each case. 2. Similarity Function: Describe and justify the similarity function for comparing image patches you are using in your programming assignment on texture synthesis. 3. Markov Property: Describe the Markov Property in terms of its application in texture synthesis. What are the pros and cons of relying on this property? (hint: how can one be faithful to near-regular textures during texture synthesis?) 4. Marching Squares/Cubes: Describe the marching squares / cubes method for contour surface representation. Which two of the following congurations are redundant? Explain how you nd it.

5. Camera alignment in 3D: Given a camera with eye position e, up vector u, and look-at vector v. Write the sequence of matrices that transforms the world (XYZ with origin o) to align the camera described with the default camera position and orientation. Assuming that the default camera position and orientation has the eye at (2,3,4), upvector y, and looks in the x direction. Suggestion: draw and label the respective coordinate systems to illustrate your answer. 6. Extra Credit: Evaluation of NRT synthesis results Jamaican dancers (gender, dance ability) Extra credits are also listed below for the 4th programming assignment below: 1. NRT synthesis: use an existing lattice detection algorithm to nd the lattice and develop an algorithm that respects the detected regularity of the input texture. References: NRT IJCV paper 2005: The Promise and Perils of Near-Regular Texture 2

NRT SIGGRAPH paper 2004: Near-Regular Texture Analysis and Manipulation PAMI paper and algorithm 2009: Deformed Lattice Detection in RealWorld Images using Mean-Shift Belief Propagation PAPERS AVAILABLE AT: http://vision.cse.psu.edu/publications/publications.shtml MATLAB CODE AT: http://web.me.com/mipark/MPark2/Code 2. Image defencing: Given the PAMI 2009 lattice detection (see above) + fence mask extraction code, develop an inpainting algorithm to remove the fence from the image. Compare yours against Photoshop CS5s content-aware ll tool. Reference: CVPR 2008: Image De-fencing, ACCV 2010: Image Defencing Revisited (papers at http://vision.cse.psu.edu/publications/publications.shtml) 3. Dance texture: Given the mapping between FK (forward kinematics) and dancer textures, and learned discriminative joint features, construct a good similarity function for dance texture synthesis.

S-ar putea să vă placă și