This paper deals with the problem of three-dimensional volume reconstruction from two-dimensional projections with unknown orientations. This situation occurs in cryo-electron microscopy. A method determining the orientations and reconstructing the volume jointly from scratch (ab initio) is presented in this work. It is based on a heuristic cost minimization including a comparison of the input projections with the projections obtained from the image being reconstructed and the orientations information and a data fitting based on the common-lines. The method is tested on synthetic data and compared to SIMPLE software on real data.