Usually

camera self-calibration is nonlinear. A linear camera calibration technique based on four point correspondences is proposed in this paper in order to obtain a linear method

and a robust Euclidean reconstruction under the random sample consensus (RANSAC) algorithms is computed from the images. The main results are: Assuming the camera undergoes two sets of translations

of which the corresponding rotation axes of camera are not parallel

then Euclidean geometry of a scene can be reconstructed linearly from four point correspondences. Experimental results on synthetic and real images have validated the proposed linear technique.