Biomedical Imaging Group
EPFL   Student Projects: Christophe Magnard
English only    BIG > Teaching > Student Projects > Christophe Magnard
Home page
Tutorials and Reviews
Download Algorithms
bulletStudent projects


Automatic calibration of paracatadioptric camera

Christophe Magnard, Microengineering Section

A panoramic camera is composed of a camera and a catadioptric system, i.e., a system of lenses and mirrors. Such a system allows enhancing the field of view while preserving a single viewpoint.

The acquired images are deformed and hence one needs to unwarp them. To unwarp, we need to know the parameters of the system. These are recovered through a calibration process. Usually, the calibration is done manually using the external circle of the image and corresponding values provided by the manufacturer.

This project consists of automatically calibrating a panoramic camera that uses a parabolic mirror, employing the contents of the acquired image.

Set up of the camera

The proposed calibration method uses the properties of the projection of straight lines in real world onto the image plane. These straight lines are projected onto circles, if they are not vertical and onto straight lines pointing to the focus, if they are vertical. The properties of these straight lines and circles allow determining the parameters of the paracatadioptric system.

The first part of this work is the algorithm for detecting the parameters. This algorithm first detects the contours via an edge detection and a robust post processing, and fits circles and straight lines on these contours. From these circles and straight lines, the required parameters are detected.

Graphical result of the calibration

The second part consists of recovering the perspective image from the acquired image. An interface is provided that allows to specify the perspective image. This is done through a Java application.

Perspective image reconstructed