Research Projects:

Video-based Crowd Synthesis	[ Ph.D. Dissertation | video ]
Target Segmentation and Tracking	[ video, 10.8MB, AVI ]
Human Video Textures	[ video, 74MB AVI, DivX codec ]
Projector-Guided Painting	[ video, 53MB, Quicktime ]
Virtual Rear Projection	[ video 1, 10.8MB, MPEG4, video 2, 26.8MB ]
Multi-Planar Projected Displays	[ video, 6.3MB, DivX ]
Real-Time Projector Calibration	[ video, 13MB, MPEG4 ]

Publications:

Journal Articles
	Video-based Crowd Synthesis Matthew Flagg and James M. Rehg IEEE Transactions on Visualization and Computer Graphics (TVCG), online pre-print published on November 30, 2012. [ pdf, 17.7MB ]
	Motion Coherent Tracking using Multi-Label MRF Optimization David Tsai, Matthew Flagg, Atsushi Nakazawa, James M. Rehg International Journal of Computer Vision (IJCV), November 2012, Volume 100, Issue 2, pp 190-202. [ pdf, 8.4MB ]
	Shadow Elimination and Blinding Light Suppression for Interactive Projected Displays Jay Summet, Matthew Flagg, Tat-Jen Cham, James M. Rehg, and Rahul Sukthankar IEEE Transactions on Visualization and Computer Graphics (TVCG), Volume 13, Number 3, May/June 2007. [ pdf, 0.37MB ]
Conference Papers
	Motion Coherent Tracking with Multi-Label MRF Optimization David Tsai, Matthew Flagg and James M. Rehg, to appear in British Machine Vision Conference (BMVC) for ORAL presentation. Aberystwyth, UK, August 2010. [ pdf, 8.9MB | project website | Winner of Best Student Paper Award ]
	Human Video Textures Matthew Flagg, Atsushi Nakazawa, Qiushuang Zhang, Sing Bing Kang, Young Kee Ryu, Irfan Essa, and James M. Rehg, to appear in Symposium on Interactive 3D Graphics and Games (I3D). Boston, MA, February 2009. [ pdf, 24MB | project page | video, 74MB AVI | featured on the back cover ]
	Projector-Guided Painting Matthew Flagg and James M. Rehg, Appeared in Proceedings of User Interface Software and Technology (UIST). Montreux, Switzerland, October 2006. [ pdf, 2.2MB, project page, PPT slides, 39MB, talk video, featured on the cover! ]
	GVU-PROCAMS: Enabling Novel Projected Interfaces Jay Summet, Matthew Flagg, James M. Rehg and Gregory Abowd Appeared in ACM Multimedia 2006. Santa Barbara, California, October 2006. [ pdf, 3.5MB ]
	A Flexible Projector-Camera System for Multi-Planar Displays Mark Ashdown, Matthew Flagg, Rahul Sukthankar, James M. Rehg, Appeared in IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2004), Jun. 2004. [ pdf, 0.67MB ]
	Projected Light Displays Using Visual Feedback J. M. Rehg, M. Flagg, T.-J. Cham, R. Sukthankar, and G. Sukthankar, Appeared in Intl. Conf. on Control, Automation, Robotics, and Vision, Singapore, Dec. 2-5, 2002. [ pdf, 0.56MB ]
Workshop Publications
	Computational Support for Creativity via Capture and Access Matthew Flagg, Christopher Skeels, James M. Rehg, To appear in Computational Creativity Support held in conjunction with ACM CHI 2009, April, 2009, Boston, MA, USA. [ pdf, 1.5MB ]
	Improving the Speed of Virtual Rear Projection: A GPU-Centric Approach Matthew Flagg, Jay Summet, James M. Rehg, Appeared in IEEE International Workshop on Projector-Camera Systems (PROCAMS) held in conjunction with the IEEE International Conference on Computer Vision and Pattern Recognition (CVPR 2005), June, 2005, San Diego, CA, USA. [ pdf, 1.2MB, video, 10.8MB, DiVX ]
	Robust Projected Displays for Ubiquitous Computing Jay Summet, Matthew Flagg, Mark Ashdown, Rahul Sukthankar, James M. Rehg, Gregory Abowd, Tat-Jen Cham Appeared in Proceedings of Ubicomp Workshop on Ubiquitous Display Environments, September 2004. [ pdf, 0.37MB ]
	Increasing the Usability of Virtual Rear Projection Jay Summet, Matthew Flagg, James M. Rehg, Gregory M. Corso, Gregory D. Abowd, Appeared in IEEE International Workshop on Projector-Camera Systems (PROCAMS) held in conjunction with the International Conference on Computer Vision (ICCV 2003), Nice, France, October 2003. [ pdf, 0.70MB ]
Tech Reports
	Shadow Elimination and Blinding Light Suppression for Interactive Projected Displays Jay Summet, Matthew Flagg, Tat-Jen Cham, James M. Rehg, Rahul Sukthankar, Technical Report GIT-GVU-06-02 [ pdf, 1.2MB ]
	Oil Painting Assistance Using Projected Light: Bridging the Gap Between Digital and Physical Art Matthew Flagg and James M. Rehg, Technical Report GIT-GVU-05-35 [ pdf, 1.6MB ]

Virtual Rear Projection

Jay Summet and I are working on a system that leverages two projectors to compensate for shadows and blinding light on users - we call this virtual rear projection because the goal is to emulate a rear projection display using front projection.

Watch this video (6.5MB, DIVX avi) of our demonstration, given at ICCV 2003 in Nice, France. Here is a 1 page demonstration explanation (PDF).

In the following publication, we report the results of a user study and the use of markers to calibrate the system in real-time with the combined effort of increasing the usability of virtual rear projection:

Jay Summet and I gave a GVU brown bag talk on Virtual Rear Projection on March 18, 2004. You may watch our presentation in streaming Quicktime and download slides.

Virtual rear projection requires calibration for geometric alignment of multiple projected displays and photometric compensation for illumination differences caused by reflectance properties of the screen and variations due to physical projector placement. Here are two images that illustrate these needs:

Projector-Camera System for Multi-Planar Displays

I helped create a system that calibrates a projected display onto multiple planes. This system works by discovering the planar surfaces using a form of structured light and calculating a set of metric-rectified homographies. A metric-rectified homography is a mapping from the projector to the plane that is precise enough to let someone make metric measurements from each plane, such as distances on a map for both a plan view and topographic view. Here is a technical report on this system:

Real-Time Projector Calibration

I work with projectors and cameras to automatically correct keystone distortions. With our system, the user may put four blue poker chips on a wall inside a region of projected light and the sytem will automatically project inside the region defined by the poker chips. Furthermore, the system will account for environmental disturbances such as movements of the projector, camera, or poker chips. To learn more about this system, read the ICARCV publication: