I   Video Browsing

• Preferred experience: Some understanding of image processing.
  
• Project description: Commonly lectures are video taped, and archived. Archives of these lectures, might contains many thousands of hours of recorded lectures. Searching specific subjects in these videos is a tedious task. There are many methods to fasten the search, but much more work is required.
  We collaborate with IBM research center to propose new methods for searching videos. We need students help in implementing and testing several algorithms that we developed.
  
• Collaborators: IBM research group and Prof. Kobus Barnard (UofA).

• Preferred experience: Image processing, algorithms, computer graphics.
  
• Project description: The majority of cancer patients receive radiation therapy as part of their treatment. Radiation therapy works by directing radiation beams at the targeted areas from different directions, but it also results in the irradiation of normal organs which can only tolerate a certain dose of radiation before injury occurs. The goal of radiotherapy is to maximize the tumor dose while minimizing dosing of normal tissues to achieve the best survival with the fewest complications.
  
  Typically, a CT scan of the patient is obtained to image the tumor and to plan the radiation fields. Patients receive daily doses of radiation over a period of several weeks, requiring precise placement of the patient in the treatment unit each day. The patient positioning is performed based on the patient's skin; however, the current methods which rely on the alignment of body marks or tattoos to reference laser lines are only accurate to 5 mm to 10 mm. In a join work with the radiation oncology department at the University medical center, we are designing a body surface sensing system which will form a real time three-dimensional model of the skin of the patient and determine the location of the patient relative to the radiation beam. The system will compare this model with the patient model obtained from the CT scan to determine if the alignment of the patient agrees with the planned alignment and will direct the therapist to position the patient properly. The accuracy reached by this system is expected to be better than 2 mm. This will permit smaller margins on the radiation beams resulting in better treatment outcomes.
• Collaborators: Prof. Russell Hamilton from the radiation oncology center at UMC, and Prof. Kobus Barnard(UofA).

• Preferred experience: Image processing, databases.
• Project Description: We are involved in the LSST at which a large telescope will survive the sky. The images taken by the telescope would be used for many researchers purposes. The goal of the CS department at UofA is to find moving objects (e.g. asteroids) to find their trajectories, and decide whether they have been observed before. This involves both manipulating several large data bases, and facilitating graph matching algorithms.
  
• Collaborators: Prof. Bongki Moon (UofA), Prof. Kobus Barnard (UofA) and many other.

• Preferred experience: Some understanding of image processing. Geometric Algorithms implementation. Geographic Information Systems - a plus. Familiarity with fragments programming is also a plus
  
• Preferred experience: Assume we measured the amount of rain falling at several points, at which rain gauges are installed. We need to smartly guess what is the amount of rain falling at other locations at which we do not have rain gouges. The process of evaluating the value of a function (rain fall in our example) at a query point q from readings of the function at other smaple points where its value is known, is called interpolation. There are numerous methods for interpolation. One which is of particular interest is the Area-Stealing interpolation (also known as Natural Neighboring interpolation. This methods enjoys many properties, which made it popular between researchers of many fields. However, once the number of sample points and query points are large (which is typically the case for physics experiments, Geographic Information Systems (GIS) and other realistic cases), this methods requires many amount of computing resources. We developed several algorithms for fasten this methods. We obtained nice results that were published as preliminary work. Other algorithms are awaiting to be implemented and tested.
  
• Collaborators: Dr. Cesim Erten

V   Sensor location

• Preferred experience: Geometric Algorithms implementation
  
• Project description: We have develped several algorithms for the problem of how to finds optimal locations of sensors (e.g. cameras) that collaboratively can see or "guard" an area such as a a building or a or a mountain ridge. This type of problems have many variant, such as we need to maximize the area that the sensors "see", whether the sensors are fixed, or can move within the area (e.g. mountain on robots), what is the communication models between the sensors and between the sensors to the base-station(s), etc. We are currently experimenting several solutions to these problems, and more work is needed.
  
• Collaborators: Prof. Joseph S.B. Mitchell, Dr. Cesim Erten, Prof. Srinivasan Ramasubramanian and Prof. Stephen Kobourov .