Bundle adjustment is one of the cornerstone to recover the scene structure from a sequence of images. The main drawback of this technique, due to nonlinear optimisation, is the need of initial conditions for intrinsic and extrinsic camera parameters and for the 3D structure that we want to reconstruct. In this paper, we demonstrate that we can reduce, comparatively to the standard approach, the number of degrees of freedom. Synthetic and real experiments show that our method is as accurate as the classical one while reducing the computation time and increasing widely the convergence domain on 3D shape and intrinsic parameters.