Maximum-likelihood (ML) image refinement is a promising candidate to improve attainable resolution limits in 3D-EM. However, its large CPU requirements may prohibit application to 3D-structure optimization. We speeded up ML image refinement by reducing its search space over the alignment parameters. Application of this reduced-search approach to a cryo-EM dataset yielded practically identical results as the original approach, but in approximately one day instead of one week of CPU. This work has been implemented in the public domain package Xmipp. Documentation and download instructions may be found at: http://www.cnb.uam.es/~bioinfo