| The process of creating 3D models from data captured from real objects is called 3D reconstruction. Given images of a set of 3D points, with known correspondence among them, we would like to reconstruct the coordinates of the 3D points and discover the perspective projection matrices associated with the cameras used to capture these images. Once this information is obtained, we would be able to compute any novel view of these 3D points from a virtual camera placed at any position and orientation. The analytical method which estimates the perspective projection matrices has the limitation that the reconstructed models are ambiguous up to a projective transformation. To overcome this, we estimate the camera intrinsic and extrinsic parameters directly. We solve this problem as an optimization problem using the Particle Swarm Optimization (PSO) algorithm. |
