Stereo vision is a part of the broad field Computer vision. It deals with the three-dimensional reconstruction of a scene. Therefore usually two cameras are mounted in parallel and capturing simultaneously. Furthermore, correspondences are searched between the different views using a stereo matching method in order to calculate depth by triangulation. This thesis is about a special stereo vision system, a projector-camera sensor. Unlike classical stereo vision sensors where two cameras are used, one camera is replaced by a pattern projector. The three dimensional reconstruction is based on the same algorithms and processes. The only difference is, that not two camera images are used for the calculations but only one where the projected pattern is visible and in addition the pattern which is projected. First, such a sensor has to be calibrated accurately. Since only one camera is available, the calibration method differs slightly from the classical stereo calibration procedure. In this thesis a projector-camera calibration method is presented where a precalibrated camera is needed to calculate all the intrinsic and extrinsic calibration parameters. If this information is available the rectification and stereo matching procedure can be performed and furthermore, depth can be computed by triangulation. The stereo matching method that is discussed here, is a pixel-based approach called census. Furthermore, an extension of a classical two-camera stereo system by a pattern projector is presented. This special sensor, consisting of two cameras and a projector is therefor a combination of a two-camera system and a projector-camera system. The purpose of this combined sensor is to reduce the reconstruction difficulties of both individual systems to result in a denser combined depth map. Next, the implemented software, including the projector-camera calibration, the rectification and stereo matching in MATLAB as well as the extensions of the existing High Speed Stereo Engine S3E developed by the Austrian Institute of Technology, is documented. Especially for the projector-camera calibration an automatic approach is developed and presented in this thesis which is based on the feature detection algorithm SIFT. Finally, the projector-camera system is tested and verified. For that purpose, real-world scenes are measured and compared with the calculated results. Additionally the matching quality is analysed and it is shown that the best results are achieved by using a blurred version of the projected pattern for the stereo matching calculations. Furthermore, the depth maps of the projector-camera system are compared with the maps of the two-camera system as well as with the maps of the combined sensor.