One of the major goals of computer vision is the recovery of spatial information about the environment. Euclidean reconstruction can be done using images of a scene taken by a camera and Euclidean reconstruction requires that the camera calibration is known. Classical approaches of the camera calibration assume that the cameras are calibrated beforehand, but a great interest in uncalibrated vision and on.line calibration has arisen during the last couple of years. For many practical applications, it is important to relax the self.calibration conditions to allow for changing internal camera parameters.
In this thesis, we implement an algorithm to extract 3-D information of the scene in the Euclidean space for changing internal camera parameters and present new method to find initial values that are needed in nonlinear minimization process. But this algorithm is very sensitive to the input noise (i.e. noise from corresponding points between images). We show the reason why this algorithm is sensitive to the input noise by experiments. And we propose a new constraint that makes an algorithm robust to the input noise.