Camera calibration is necessary to obtain 3D information from 2D images of the scene. There are different techniques which are based on photogrammetry or self-calibration. Photogrammetric techniques take an image of a known scene which has a three, two or one-dimensional template. With the coordinates of the points in the template and its correspondences in the image the camera is calibrated. Self-calibration techniques take several images of the scene and using the rigidity of the scene they fix restrictions to resolve the camera calibration.  
As a result of the calibration the camera intrinsic and extrinsic parameters are computed. Computing the parameters is not an exact process because of the uncertainties. This uncertainty came from constructive imperfections and mechanical misalignments of the lenses or sensor, and because of the image is processed to obtain points positions within them. Results depend so much of the calibration template as the calibration algorithm and also on processing of the data before resolving the camera calibration. A lot of techniques exist to resolve it. 
Since it is impossible to compute the exact value of the camera parameter, it proves to be very interesting to compute an interval in which the parameter is. The camera parameter together with its uncertainty improves all procedures in which it participates such as 3D scene reconstruction. By the other hand, when a camera calibration is going to be carried out, a lot of questions arise about the number of points to put into the calibration template, number of images to take of the template and also from where and how to orient the camera to take the images.
This work gives the answer to all these questions. First, the best performance camera calibration algorithm is chosen. A previous data processing has been added to this calibration method to improve the results. Data processing allows to reduce the noise of the measurements and also to associate the uncertainty to each data. If each data has an individual uncertainty, it can be use afterwards in the calibration process according with its uncertainty. As a result, camera parameters are computed together with its uncertainty also. To resolve the questions about the number of points in the template, or number of images of it, the camera calibration process has been characterized from a statistical point of view. This allows defining how the results are affected by the uncertainty of the input data and therefore, the number of images and points can be determined in order to obtain the best performance of the calibration algorithm. Finally, the optimal positions and orientations of the camera to take the images are defined. These conditions do not depend on the camera features and therefore they can be used with in any camera calibration process.