3D Reconstruction Based on Pseudo-Linearization and Errors-in-Variables Model

This paper proposes a general approach of 3D reconstruction, a major problem arising in computer vision and virtual reality, based on a combination of Pseudo-Linearization and Errors-in-Variables model. The proposed approach concerns a bunch of corrupted measurements under nonlinear constraints, and optimizes the estimation by taking errors into account. Furthermore, we set a synthetic projective model and adopt a standard deviation-expectation criterion to evaluate the performance or our method applied in 3D reconstruction. Also, some test images are picked from an image database to give this method a chance to demonstrate its performance in our experiments. Finally, as a successful application, this method is used in a calibration-free augmented reality system.