UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Retrieving camera parameters from real video images Li, Ning

Abstract

The motivation behind Computer Augmented Reality (CAR) is to effectively merge real video images and computer generated images to enhance the usefulness of the two sources of information. To fulfill this goal, we need to solve many related tasks which need expertise both in computer graphics and computer vision. This thesis is focused on one important aspect of CAR: establishing common viewing conditions between real video and computer generated images, that is, recovering the camera extrinsic parameters (position and orientation) as well as camera intrinsic parameters from real video images, so that the computer generated objects can be inserted accordingly, and the combined environment can be similarly rendered. In this thesis, we assume that the camera is moving when the real video images are taken, and we know nothing about the camera intrinsic parameters - the camera is uncalibrated. First a corner detector is used to acquire some feature points in the image. With the knowledge of Euclidean 3D measurements for six or more feature points in the image, the matrix which transforms 3D coordinates into 2D coordinates can be acquired fairly accurately, then the camera parameters can be recovered from this matrix. A system is built which makes the whole process work as automatically as possible. We tested our system with both synthetic and real images. We found that the system is able to produce useful results in most cases. When applied in the context of CAR to insert computer generated objects into real images, the results provide useful information about the real camera, so the synthetic camera can be manipulated accordingly.

Item Media

Item Citations and Data

Rights

For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.