I recently wrote a new textbook called Computer Vision for Visual Effects, which was published by Cambridge University Press in Fall 2012. The book describes classical computer vision algorithms used on a regular basis in Hollywood (such as blue-screen matting, structure from motion, optical flow, and feature tracking) and exciting recent developments that form the basis for future effects (such as natural image matting, multi-image compositing, image retargeting, and view synthesis). It also discusses the technologies behind motion capture and three-dimensional data acquisition. More than 200 original images demonstrating principles, algorithms, and results, along with in-depth interviews with Hollywood visual effects artists, tie the mathematical concepts to real-world filmmaking. Check it out!
Video Analytics in Camera Networks
I am particularly interested in computer vision problems that occur in large networks of cameras dispersed throughout an environment. Several years ago, an NSF CAREER award supported my lab's investigation into distributed solutions for determining visual overlap and camera calibration in large dynamic camera networks. These days, my main interest in this area is the design of video analytics algorithms applied to network camera images, like you might find at an airport. We're particularly interested in problems involving counterflow and re-identification. This video gives a brief overview. Our research in this area is currently supported by ALERT, the DHS Center of Excellence on Explosives Detection, Mitigation and Response.
I lead the Controls Thrust in the NSF Engineering Research Center for Smart Lighting, a multi-university effort headquartered at RPI. Our group generally studies how the input from a distributed network of multimodal sensors can drive advanced control algorithms and color-tunable LED lights to achieve a desired light field in a room. In particular, my group studies how time-of-flight sensors mounted in the ceiling of a smart room can locate occupants in real time, and how graphical simulation can be used to help pre-visualize and tune the parameters of lighting control algorithms before a system is installed.
Computer Vision, Machine Learning, and Optimization for IMRT
Through my affiliation with CenSSIS, I supported several undergraduate and graduate projects in biomedical image processing. I am particularly interested in computer vision and machine learning problems related to intensity modulated radiotherapy (IMRT), an exciting new technology for cancer treatment. One project involves developing computer vision algorithms to aid in the automatic segmentation of organs from 3-D CT scans acquired immediately prior to radiation treatment. A second project, now supported by an NIH R01 award, investigates the relationship between a patient's body/organ geometry and the multiple radiation beams that are used to treat their cancer. Our algorithm, called Reduced Order Constrained Optimization, or ROCO, creates clinically acceptable IMRT plans in a matter of minutes for several treatment sites, including the prostate, lung, and nasopharynx. This work proceeds in close collaboration with medical physicists at Memorial Sloan-Kettering Cancer Center.
Integrating LiDAR and Digital Images
I am very interested in LiDAR (Light Detection and Ranging), a laser range scanning technology that allows us to acquire detailed 3D models of real-world environments. Our research in this area includes keypoint detection, registration, integration of range and visual imagery, and probabilistic object detection, and has been sponsored by the US Army Intelligence and Security Command and DARPA. We are in the process of building an accurate 3D model of the RPI campus with our high-quality scanning LiDAR system.
Change Detection and Understanding
I am interested in change detection as well as change understanding in image and video sequences. Several years ago, we undertook a comprehensive survey of pixel-level change detection algorithms. We are more broadly interested in leveraging pixel-level change detection algorithms, along with domain-specific models for objects and behaviors of interest, to produce semantic change understanding algorithms that can help interpret and annotate image sequences the same way an expert observer would. To date, we have demonstrated this capability in the context of biomedical image sequences (e.g. time-lapse video of neurons and stem cells) to quickly and accurately summarize megabytes of image sequence data.
I joined the Electrical, Computer, and Systems Engineering department at Rensselaer Polytechnic Institute in August, 2001, where I am now a Full Professor. I have a dual B.A. degree in math and computational and applied math from Rice University, an M.A. in computational and applied math from Rice University, and M.A. and Ph.D. degrees in electrical engineering from Princeton University. I was an intern at the Mathworks, developing numerical linear algebra and signal processing routines. During my Ph.D. I investigated several estimation problems in digital video, including the efficient estimation of projective transformations and the synthesis of photorealistic "virtual video", in collaboration with IBM's Tokyo Research Laboratory.
My current research interests include computer vision problems related to modeling 3D environments with visual and range imagery, designing and analyzing large camera networks, and machine learning problems for radiotherapy applications. I am affiliated with the NSF Engineering Research Center for Subsurface Sensing and Imaging Systems (CenSSIS), the DHS Center of Excellence on Explosives Detection, Mitigation and Response (ALERT), and Rensselaer's Experimental Media and Performing Arts Center (EMPAC). I received an NSF CAREER award in March 2003 and was a member of the 2007 DARPA Computer Science Study Group. I am a Senior Member of the IEEE and an Associate Editor of IEEE Transactions on Image Processing. In Fall 2012, Cambridge University Press published my textbook Computer Vision for Visual Effects.
Here is a brief Curriculum Vita.
Indrani Bhattacharya, Ph.D.
Dan Kruse, Ph.D. (co-advised with John Wen)
Srikrishna Karanam, M.S.
Austin Li, Ph.D.
Omar Al-Kofahi, Ph.D. (2005, co-advised with Badri Roysam; now with American Science and Engineering)
Eric Ameres, M.S. (2011, now with EMPAC)
Srinivas Andra, M.S. (2003)
Shaimaa Bakr, M.S. (2014)
Jacob Becker, M.S. (2009, co-advised with Chuck Stewart; now with Kitware)
Andrew Calcutt, M.S. (2010)
Siqi Chen, Ph.D. (2011, now with Netflix)
David Doria, Ph.D. (2012, now with Army Research Lab)
Zhaolin Cheng, M.S. (2006, now with Captira Analytics)
Anil Cheriyadat, Ph.D. (2009, now with Oak Ridge National Labs)
Haeyong Chung, M.S. (2005)
Dhanya Devarajan, Ph.D. (2006, now with Immersive Media, Portland, OR)
Yongwon Jeong, Ph.D. (2006; now with Pusan National University, South Korea)
Li Jia, Ph.D. (2014)
Chao Ling, M.S. (2006, co-advised with Paul Schoch)
Renzhi Lu, Ph.D. (2007, now with Morgan Stanley, NYC)
Linda Rivera, Ph.D. (2013, now with Pratt and Whitney, East Hartford, CT.)
Eric Smith, Ph.D. (2011, co-advised with Chuck Stewart; now with Kitware)
Ziyan Wu, Ph.D. (2014, now with Siemens Corporate Research, Princeton, NJ)
For Prospective StudentsI receive many e-mails from prospective students asking me if I am hiring students into my group or if I will consider their applications based on an attached resume. I unfortunately cannot respond to each of these individually. You should know that I do not make decisions regarding admission or financial aid, and I only consider prospective students who have gone through the formal application process to our department. If you refer to me in your statement of purpose as someone with whom you are interested in working, I will review your application for a possible fit with my group. I hire students with strong image processing and computer vision backgrounds; if your background is in computer hardware or networking, I am probably not an appropriate advisor for you. I am much more likely to respond to your e-mail if you show that you are actually well-informed about my research (e.g., have read some of my papers) instead of expressing a generic interest in computer vision.