Events - Colloquia & Seminars

CCIS Colloquium Spring 2006

Separating Foreground from Background in Video, and Other Applications of a Single-axis, Multi-sensor Camera

Speaker: Morgan McGuire (Brown University)

Date: Friday, February 17, 2006

Talk: 12:00 pm, 366 WVH

Abstract

A television weatherforecaster appears with a satellite map but is really standing in front of a blue screen. \emph{Matting} is the process of separating the forecaster from the screen; doing this if the forecaster is instead in front of a complex moving background, say traffic in New York City, is an open problem. I present a solution to this problem as one instance of what can be accomplished with a camera containing multiple video sensors.

Most cameras contain a single sensor that measures, at each pixel, the amount of light near specific red, green, and blue wavelengths. Using multiple sensors we can measure more wavelengths, multiple exposures, different polarizations; we can even consider different lens parameters.

Future cameras will likely have these capabilities, which raises interesting questions:

• What are good designs for multi-sensor cameras?
• How can we use them to solve current problems?
• Given multiple sensors, what new applications can we explore?

In this talk, I present a general framework for working with multi-sensor cameras, demonstrate several applications, and describe methods for construction, calibration, and optimizing optical design.