|Correction of optical deformations for a panoramic camera|
Section Microtechnique, EPFL
|Wide field camera systems have many interesting applications, e.g. surveillance for security applications, videoconference or robot navigation. A wide field of view is generally achieved by using several cameras or a mechanically moving camera.|
The acquisition system used in this project consists of a conventional CCD camera in front of a convex parabolic mirror. This minimizes cost and complexity of the system. The mirror extends the field of view to 360 degrees but the captured image is strongly deformed. This makes interpretation of the image difficult.
The purpose of this project is to unwarp the deformed image to perspective images. We derived an optical model of the system and developed a program that uses the model to unwarp the images. Given parameters specifying a perspective image, the program chooses the corresponding region of interest in the deformed image, which is then transformed to a perspective image using diffrent interpolation methods (Nearest neighborhood, Bilinear and Cubic spline).
Furthermore, the application of the Multigrid Image Reconstruction Algorithm, which has been developed at BIG, is investigated for this kind of problem. This algorithm allows reconstructing images from non-uniform distributed pixels.