Image stitching and background subtraction for moving cameras

Spring 2012

Debjani Biswas
Martial Hebert
Project description

Active structured light sensors infer the 3D shape of objects by analyzing snapshots of projected light patterns. The quality of the result depends on a clean image of the projected light pattern, i.e. with little or no background. Outside in bright sunlight the light from the sun can produce a strong background that can only be removed by background subtraction. This is especially challenging when the camera is moving. The goal of this project is to infer the background in the image with the projected light pattern. Images are taken before and after the image without projected light, stitched together into one large background image and then registered with the image with the projected light pattern. The tools to do this will be built from the openCV image library. I will install openCV, familiarize myself with the software, write the background-subtraction application and evaluate the application.

Return to project list