Perturbation analysis applied to non homogeneous camera calibration

Camera calibration process is very sensitive to noise. Errors' coming from noisy measurements of the points' coordinates, their situation in the world and the calibration algorithm itself creates a deviated solution from the real one. Therefore, it is more interesting to obtain the standard deviation and the mean value of a parameter, than only a value. This value is always false because of the input data is corrupted with noise. This standard deviation represents the quality of the computed parameter. In this paper a method to compute this standard deviation of the projection matrix elements is presented. In this case, the projection matrix is computed with the non homogeneous solution based on the least squares technique.