In this paper one of the important tasks of modern computer geometry is considered: creating effective algorithms for gluing different flat images of the same object. Images are obtained by central projection from different points of view. We use numerical simulation for comparison of three known algorithms for gluing---simple linear algorithm, normalized linear algorithm and direct algorithm. In each case stability to perturbations of the initial data and speed of calculations were estimated. The results confirm hypothesis of G.V. Nosovski\'\i and E.S. Skripka that the direct algorithm proposed in their work [Error estimation for the direct algorithm of projective mapping calculation in multiple view geometry, Proceedings of the Conference ``Contemporary Geometry and Related Topics'', Belgrade, Serbia-Montenegro, June 26--July 2, 2005, Faculty of Mathematics, University of Belgrade, 2006, pp.~399--408] is the most accurate and fast one.