In this dissertation, new depth camera calibration techniques were developed using only depth images and they are adopted to the case of multiple depth cameras. In most studies, depth camera calibration was performed using infrared images of checkerboard and obtained parameters were accepted as depth camera parameters. Since this approach utilizes readily available and commonly used calibration methods, it can be considered as a practical solution. However it does not take advantage of the depth measurements and leads to calibration errors that cause misregistration in 3D. Even if additional complementary methods such as iterative closest point method are applied, misregistration artifacts cannot be eliminated. As a solution, a 3-dimensional tetrahedron calibration object with 4 equiradius spheres was designed and an ellipse-based method was developed to detect the sphere centers on the depth image plane. By using the sphere centers and the depth values with the depth error removed, the well-known projection matrix method has been modified to require 4 matching points instead of 6 as in the original method, which works efficiently with the proposed setup. To estimate the parameters of four depth cameras together, two new parameter estimation methods have been developed that include depth values in the traditional bundle adjustment (BA) and calculate the reprojection error in 3D. Compared to the calibration using the IR images of the checkerboard, the average error resulting from proposed methods was reduced significantly. Since the developed methods are independent of the technology of depth cameras, they can work for all depth cameras.