The approach developed in this thesis, allows to determine the unknown camera position from a given picture of an object. It is based on the hierarchically iterative registration. Therefore the picture is compared with rendered pictures from different camera viewpoints. The shape matching algorithm used for the comparisons is based on hu-moments. The viewpoint with the largest agreement forms the basis for the generation of new views. The error rate of the shape matching algorithm was lowered by additional comparisons of the mass centroid axis angles. The final algorithm reached an accuracy of 0,7 degrees of deviation from original camera position after only six iterations.