Projects can be done individually or in teams of two. Below is a list of possible project topics. This list is by no means exhaustive. In fact, we encourage you to come up with your own idea for a project topic.
- User study (Team of 2;): Identify 2-3 visualization techniques, implement them. Next, design and perform a user study and prepare a report on the effectiveness of the visualization techniques for the listed tasks. Reference: David H. Laidlaw, Robert M. Kirby, Cullen D. Jackson, J. Scott Davidson, Timothy S. Miller, Marco da Silva, William H. Warren, Michael J. Tarr, "Comparing 2D Vector Field Visualization Methods: A User Study," IEEE Transactions on Visualization and Computer Graphics, vol. 11, no. 1, pp. 59-70, Jan./Feb. 2005, doi:10.1109/TVCG.2005.4
- Protein visualization (Team of 1-2): Extend existing protein visualization tool to include other representations (cartoon, surface, etc.). Reference: Pranav D. Bagur, Nithin Shivashankar, and Vijay Natarajan, Improved quadric surface impostors for large bio-molecular visualization, ICVGIP, 2012. Paper and video of tool available at http://www.csa.iisc.ernet.in/~vijayn/pubs/index.html
- Graph visualization (Team of 2; multiple teams): Create a graph visualization tool by implementing existing algorithms for constructing a graph layout. Choose graph drawing algorithms depending on your choice of graphs from a particular domain (for example, social networks, biological networks, etc.) Study generic graph visualization tools such as Tulip (tulip.labri.fr/) and dot (www.graphviz.org/) for reference.
- Medical visualization (Team of 2; multiple teams): Design and develop a visualization tool with features that are specialized for medical data. Automatic identification of interesting isovalues, automated design of transfer functions, ability to make measurements, segmentation, are some possible features that could be incorporated into the tool). You could focus on a particular type of medical imaging data such as CT, MRI, PET, ultrasound, etc. Sample references: 1. Contour Spectrum by Bajaj et al. Vis 1997. 2. Semi-Automatic Generation of Transfer Functions for Direct Volume Rendering, Gordon Kindlmann and James W. Durkin, Vis 1998. 3. Multidimensional transfer functions in volume rendering of medical datasets by Tor Øyvind Fluør, Masters Thesis, University of Oslo.
- 2D transfer functions (Team of 1-2): Implement a 2D transfer function and apply it to achieve better quality volume rendering. References: 1. Multi-Dimensional Transfer Functions for Interactive Volume Rendering, Joe Kniss, Gordon Kindlmann, and Charles Hansen, 2002. 2. Cheuk Yiu Ip, Amitabh Varshney, and Joseph JáJá, Hierarchical Exploration of Volumes Using Multilevel Segmentation of the Intensity-Gradient Histograms, Vis 2012.
- Fast extraction of isosurfaces (Team of 2): Implement 2-3 algorithms for fast extraction of isosurfaces and compare their performance. Implement optimizations that allow isosurfaces to be extracted efficiently from very large data sets.
- Isosurface extraction + level set segmentation (Team of 2) Implement a method that combines isosurfaces with level set segmentation to obtain meaningful visualizations of medical data. Reference:Flexible and Topologically Localized Segmentation. Johannson G, Museth K, Carr H. Eurographics / IEEE Symposium on Visualization 2007 (EuroVis 2007), 179-186 (Linköping, Sweden, May 23-25, 2007).
- Good quality isosurface meshes (Team of 2): Implement a variant of the marching cubes algorithm that extracts better quality isosurface meshes. Compare results with that of Marching Cubes.
- Dimensionality reduction (1-2): Implement ISOMAP or a similar dimensionality reduction algorithm and apply it together with other techniques like clustering/PCA for the visual analysis of some high dimensional data sets.
- Outlier detection (1-2): Implement an outlier detection method for high dimensional data and use it to remove data items before performing a clustering or MDS. Use GGobi or similar infovis framework. Reference: C. C. Aggarwal, P. Yu. Outlier Detection for High Dimensional Data. ACM SIGMOD Conference, 2001.
- Weather data visualization (1): Design an effective visualization of precipitation data that is available from weather simulations. Volume rendering algorithms and the transfer function should be designed specifically for this data. Reference: http://www.vets.ucar.edu/vg/WRF-NRCM/index.html and http://www.vapor.ucar.edu/
- Mesh partition viewer (Team of 1-2): Write a plugin to ParaView that allows the user to view a 2D/3D mesh partition. The mesh is pre-processed to create such a partition so that individual segments can be processed in parallel. The challenge is to layout the partition while avoiding visual clutter, providing spatial context, and choosing view direction to highlight significant partitions. Reference: http://www-users.cs.umn.edu/~oztekin/pmvis/